chore: ignore local identity files (#1001 ) (thanks @gerardward2007)

chore: ignore local identity files (IDENTITY.md, USER.md)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-16 09:58:21 +00:00 · 2026-01-16 09:54:42 +00:00 · 2026-01-16 09:51:27 +00:00 · 2026-01-16 09:46:12 +00:00 · 2026-01-16 09:39:12 +00:00 · 2026-01-16 09:36:45 +00:00
165 changed files with 4379 additions and 968 deletions
@@ -55,3 +55,5 @@ apps/ios/*.mobileprovision
 # Local untracked files
 .local/
 .vscode/
+IDENTITY.md
+USER.md
@@ -116,6 +116,14 @@
  - launchd PATH is minimal; ensure the app’s launch agent PATH includes standard system paths plus your pnpm bin (typically `$HOME/Library/pnpm`) so `pnpm`/`clawdbot` binaries resolve when invoked via `clawdbot-mac`.
 - For manual `clawdbot message send` messages that include `!`, use the heredoc pattern noted below to avoid the Bash tool’s escaping.

+## NPM + 1Password (publish/verify)
+- Use the 1password skill; all `op` commands must run inside a fresh tmux session.
+- Sign in: `eval "$(op signin --account my.1password.com)"` (app unlocked + integration on).
+- OTP: `op read 'op://Private/Npmjs/one-time password?attribute=otp'`.
+- Publish: `npm publish --access public --otp="<otp>"` (run from the package dir).
+- Verify without local npmrc side effects: `npm view <pkg> version --userconfig "$(mktemp)"`.
+- Kill the tmux session after publish.
+
 ## Exclamation Mark Escaping Workaround
 The Claude Code Bash tool escapes `!` to `\\!` in command arguments. When using `clawdbot message send` with messages containing exclamation marks, use heredoc syntax:

@@ -2,43 +2,71 @@

 ## 2026.1.15 (unreleased)

+### Highlights
+- Plugins: add provider auth registry + `clawdbot models auth login` for plugin-driven OAuth/API key flows.
+- Browser: improve remote CDP/Browserless support (auth passthrough, `wss` upgrade, timeouts, clearer errors).
+- Heartbeat: per-agent configuration + 24h duplicate suppression. (#980) — thanks @voidserf.
+- Security: audit warns on weak model tiers; app nodes store auth tokens encrypted (Keychain/SecurePrefs).
+
+### Breaking
+- **BREAKING:** iOS minimum version is now 18.0 to support Textual markdown rendering in native chat. (#702)
 - **BREAKING:** Microsoft Teams is now a plugin; install `@clawdbot/msteams` via `clawdbot plugins install @clawdbot/msteams`.
+
+### Changes
 - CLI: set process titles to `clawdbot-<command>` for clearer process listings.
+- CLI/macOS: sync remote SSH target/identity to config and let `gateway status` auto-infer SSH targets (ssh-config aware).
 - Heartbeat: tighten prompt guidance + suppress duplicate alerts for 24h. (#980) — thanks @voidserf.
+- Repo: ignore local identity files to avoid accidental commits. (#1001) — thanks @gerardward2007.
 - Sessions/Security: add `session.dmScope` for multi-user DM isolation and audit warnings. (#948) — thanks @Alphonse-arianee.
 - Plugins: add provider auth registry + `clawdbot models auth login` for plugin-driven OAuth/API key flows.
 - Onboarding: switch channels setup to a single-select loop with per-channel actions and disabled hints in the picker.
 - TUI: show provider/model labels for the active session and default model.
 - Heartbeat: add per-agent heartbeat configuration and multi-agent docs example.
- Fix: list model picker entries as provider/model pairs for explicit selection. (#970) — thanks @mcinteerj.
- Fix: persist `gateway.mode=local` after selecting Local run mode in `clawdbot configure`, even if no other sections are chosen.
- Daemon: fix profile-aware service label resolution (env-driven) and add coverage for launchd/systemd/schtasks. (#969) — thanks @bjesuiter.
+- UI: show gateway auth guidance + doc link on unauthorized Control UI connections.
+- Security: warn on weak model tiers (Haiku, below GPT-5, below Claude 4.5) in `clawdbot security audit`.
+- Apps: store node auth tokens encrypted (Keychain/SecurePrefs).
 - Daemon: share profile/state-dir resolution across service helpers and honor `CLAWDBOT_STATE_DIR` for Windows task scripts.
 - Docs: clarify multi-gateway rescue bot guidance. (#969) — thanks @bjesuiter.
 - Agents: add Current Date & Time system prompt section with configurable time format (auto/12/24).
- Agents: avoid false positives when logging unsupported Google tool schema keywords.
- Status: restore usage summary line for current provider when no OAuth profiles exist.
 - Tools: normalize Slack/Discord message timestamps with `timestampMs`/`timestampUtc` while keeping raw provider fields.
+- macOS: add `system.which` for prompt-free remote skill discovery (with gateway fallback to `system.run`).
 - Docs: add Date & Time guide and update prompt/timezone configuration docs.
 - Messages: debounce rapid inbound messages across channels with per-connector overrides. (#971) — thanks @juanpablodlc.
 - Messages: allow media-only sends (CLI/tool) and show Telegram voice recording status for voice notes. (#957) — thanks @rdev.
 - Auth/Status: keep auth profiles sticky per session (rotate on compaction/new), surface provider usage headers in `/status` and `clawdbot models status`, and update docs.
- Fix: guard model fallback against undefined provider/model values. (#954) — thanks @roshanasingh4.
- Fix: refactor session store updates, add chat.inject, and harden subagent cleanup flow. (#944) — thanks @tyler6204.
- Fix: clean up suspended CLI processes across backends. (#978) — thanks @Nachx639.
- Fix: parse systemd ExecStart arguments when whitespace is present. (#995) — thanks @roshanasingh4.
 - CLI: add `--json` output for `clawdbot daemon` lifecycle/install commands.
 - Memory: make `node-llama-cpp` an optional dependency (avoid Node 25 install failures) and improve local-embeddings fallback/errors.
 - Browser: add `snapshot refs=aria` (Playwright aria-ref ids) for self-resolving refs across `snapshot` → `act`.
 - Browser: `profile="chrome"` now defaults to host control and returns clearer “attach a tab” errors.
- Browser: extension mode recovers when only one tab is attached (stale targetId fallback).
 - Browser: prefer stable Chrome for auto-detect, with Brave/Edge fallbacks and updated docs. (#983) — thanks @cpojer.
- Browser: fix `tab not found` for extension relay snapshots/actions when Playwright blocks `newCDPSession` (use the single available Page).
+- Browser: increase remote CDP reachability timeouts + add `remoteCdpTimeoutMs`/`remoteCdpHandshakeTimeoutMs`.
+- Browser: preserve auth/query tokens for remote CDP endpoints and pass Basic auth for CDP HTTP/WS. (#895) — thanks @mukhtharcm.
 - Telegram: add bidirectional reaction support with configurable notifications and agent guidance. (#964) — thanks @bohdanpodvirnyi.
- Telegram: skip `message_thread_id=1` for General topic sends while keeping typing indicators. (#848) — thanks @azade-c.
+- Telegram: allow custom commands in the bot menu (merged with native; conflicts ignored). (#860) — thanks @nachoiacovino.
 - Discord: allow allowlisted guilds without channel lists to receive messages when `groupPolicy="allowlist"`. — thanks @thewilloftheshadow.
+- Discord: allow emoji/sticker uploads + channel actions in config defaults. (#870) — thanks @JDIVE.
+
+### Fixes
+- Fix: list model picker entries as provider/model pairs for explicit selection. (#970) — thanks @mcinteerj.
+- Fix: align OpenAI image-gen defaults with DALL-E 3 standard quality and document output formats. (#880) — thanks @mkbehr.
+- Fix: persist `gateway.mode=local` after selecting Local run mode in `clawdbot configure`, even if no other sections are chosen.
+- Daemon: fix profile-aware service label resolution (env-driven) and add coverage for launchd/systemd/schtasks. (#969) — thanks @bjesuiter.
+- Agents: avoid false positives when logging unsupported Google tool schema keywords.
+- Agents: skip Gemini history downgrades for google-antigravity to preserve tool calls. (#894) — thanks @mukhtharcm.
+- Status: restore usage summary line for current provider when no OAuth profiles exist.
+- Fix: guard model fallback against undefined provider/model values. (#954) — thanks @roshanasingh4.
+- Fix: refactor session store updates, add chat.inject, and harden subagent cleanup flow. (#944) — thanks @tyler6204.
+- Fix: clean up suspended CLI processes across backends. (#978) — thanks @Nachx639.
+- Fix: support MiniMax coding plan usage responses with `model_remains`/`current_interval_*` payloads.
+- Fix: suppress WhatsApp pairing replies for historical catch-up DMs on initial link. (#904)
+- Browser: extension mode recovers when only one tab is attached (stale targetId fallback).
+- Browser: fix `tab not found` for extension relay snapshots/actions when Playwright blocks `newCDPSession` (use the single available Page).
+- Browser: upgrade `ws` → `wss` when remote CDP uses `https` (fixes Browserless handshake).
+- Telegram: skip `message_thread_id=1` for General topic sends while keeping typing indicators. (#848) — thanks @azade-c.
 - Fix: sanitize user-facing error text + strip `<final>` tags across reply pipelines. (#975) — thanks @ThomsenDrake.
 - Fix: normalize pairing CLI aliases, allow extension channels, and harden Zalo webhook payload parsing. (#991) — thanks @longmaba.
+- Fix: allow local Tailscale Serve hostnames without treating tailnet clients as direct. (#885) — thanks @oswalpalash.
+- Fix: reset sessions after role-ordering conflicts to recover from consecutive user turns. (#998)

 ## 2026.1.14-1

@@ -140,6 +168,7 @@
 ### Fixes
 - Packaging: include `dist/memory/**` in the npm tarball (fixes `ERR_MODULE_NOT_FOUND` for `dist/memory/index.js`).
 - Agents: persist sub-agent registry across gateway restarts and resume announce flow safely. (#831) — thanks @roshanasingh4.
+- Agents: strip invalid Gemini thought signatures from OpenRouter history to avoid 400s. (#841, #845) — thanks @MatthieuBizien.

 ## 2026.1.12-1

@@ -119,7 +119,7 @@ dependencies {
  testImplementation("io.kotest:kotest-runner-junit5-jvm:6.0.7")
  testImplementation("io.kotest:kotest-assertions-core-jvm:6.0.7")
  testImplementation("org.robolectric:robolectric:4.16")
-  testRuntimeOnly("org.junit.vintage:junit-vintage-engine:6.0.1")
+  testRuntimeOnly("org.junit.vintage:junit-vintage-engine:6.0.2")
 }

 tasks.withType<Test>().configureEach {
@@ -212,7 +212,7 @@ class BridgeSession(
      connectWithSocket(endpoint, hello, null)
    }

-  private fun connectWithSocket(endpoint: BridgeEndpoint, hello: Hello, tls: BridgeTlsParams?) {
+  private suspend fun connectWithSocket(endpoint: BridgeEndpoint, hello: Hello, tls: BridgeTlsParams?) {
    val socket =
      createBridgeSocket(tls) { fingerprint ->
        onTlsFingerprint?.invoke(tls?.stableId ?: endpoint.stableId, fingerprint)
@@ -260,15 +260,15 @@ class BridgeSession(
        else -> throw IllegalStateException("unexpected bridge response")
      }

-        while (scope.isActive) {
-          val line = reader.readLine() ?: break
-          val frame = json.parseToJsonElement(line).asObjectOrNull() ?: continue
-          when (frame["type"].asStringOrNull()) {
-            "event" -> {
-              val event = frame["event"].asStringOrNull() ?: return@withContext
-              val payload = frame["payloadJSON"].asStringOrNull()
-              onEvent(event, payload)
-            }
+      while (scope.isActive) {
+        val line = reader.readLine() ?: break
+        val frame = json.parseToJsonElement(line).asObjectOrNull() ?: continue
+        when (frame["type"].asStringOrNull()) {
+          "event" -> {
+            val event = frame["event"].asStringOrNull() ?: continue
+            val payload = frame["payloadJSON"].asStringOrNull()
+            onEvent(event, payload)
+          }
            "ping" -> {
              val id = frame["id"].asStringOrNull() ?: ""
              conn.sendJson(buildJsonObject { put("type", JsonPrimitive("pong")); put("id", JsonPrimitive(id)) })
@@ -314,20 +314,20 @@ class BridgeSession(
                },
              )
            }
-            "invoke-res" -> {
-              // gateway->node only (ignore)
-            }
+          "invoke-res" -> {
+            // gateway->node only (ignore)
          }
        }
-      } finally {
-        currentConnection = null
-        for ((_, waiter) in pending) {
-          waiter.cancel()
-        }
-        pending.clear()
-        conn.closeQuietly()
      }
+    } finally {
+      currentConnection = null
+      for ((_, waiter) in pending) {
+        waiter.cancel()
+      }
+      pending.clear()
+      conn.closeQuietly()
    }
+  }

  private fun buildHelloJson(hello: Hello): JsonObject =
    buildJsonObject {
@@ -1,5 +1,6 @@
 package com.clawdbot.android.bridge

+import android.annotation.SuppressLint
 import java.net.Socket
 import java.security.MessageDigest
 import java.security.SecureRandom
@@ -21,6 +22,7 @@ fun createBridgeSocket(params: BridgeTlsParams?, onStore: ((String) -> Unit)? =
  if (params == null) return Socket()
  val expected = params.expectedFingerprint?.let(::normalizeFingerprint)
  val defaultTrust = defaultTrustManager()
+  @SuppressLint("CustomX509TrustManager")
  val trustManager =
    object : X509TrustManager {
      override fun checkClientTrusted(chain: Array<X509Certificate>, authType: String) {
@@ -32,11 +32,10 @@ func makeBridgeTLSOptions(_ params: BridgeTLSParams?) -> NWProtocolTLS.Options?
    sec_protocol_options_set_verify_block(
        options.securityProtocolOptions,
        { _, trust, complete in
-            guard let trust else {
-                complete(false)
-                return
-            }
-            if let cert = SecTrustGetCertificateAtIndex(trust, 0) {
+            let trustRef = sec_trust_copy_ref(trust).takeRetainedValue()
+            if let chain = SecTrustCopyCertificateChain(trustRef) as? [SecCertificate],
+               let cert = chain.first
+            {
                let data = SecCertificateCopyData(cert) as Data
                let fingerprint = sha256Hex(data)
                if let expected {
@@ -49,7 +48,7 @@ func makeBridgeTLSOptions(_ params: BridgeTLSParams?) -> NWProtocolTLS.Options?
                    return
                }
            }
-            let ok = SecTrustEvaluateWithError(trust, nil)
+            let ok = SecTrustEvaluateWithError(trustRef, nil)
            complete(ok)
        },
        DispatchQueue(label: "com.clawdbot.bridge.tls.verify"))
@@ -190,14 +190,7 @@ actor CameraController {
    }

    func listDevices() -> [CameraDeviceInfo] {
-        let types: [AVCaptureDevice.DeviceType] = [
-            .builtInWideAngleCamera,
-        ]
-        let session = AVCaptureDevice.DiscoverySession(
-            deviceTypes: types,
-            mediaType: .video,
-            position: .unspecified)
-        return session.devices.map { device in
+        return Self.discoverVideoDevices().map { device in
            CameraDeviceInfo(
                id: device.uniqueID,
                name: device.localizedName,
@@ -232,7 +225,7 @@ actor CameraController {
        deviceId: String?) -> AVCaptureDevice?
    {
        if let deviceId, !deviceId.isEmpty {
-            if let match = AVCaptureDevice.devices(for: .video).first(where: { $0.uniqueID == deviceId }) {
+            if let match = Self.discoverVideoDevices().first(where: { $0.uniqueID == deviceId }) {
                return match
            }
        }
@@ -252,6 +245,24 @@ actor CameraController {
        }
    }

+    private nonisolated static func discoverVideoDevices() -> [AVCaptureDevice] {
+        let types: [AVCaptureDevice.DeviceType] = [
+            .builtInWideAngleCamera,
+            .builtInUltraWideCamera,
+            .builtInTelephotoCamera,
+            .builtInDualCamera,
+            .builtInDualWideCamera,
+            .builtInTripleCamera,
+            .builtInTrueDepthCamera,
+            .builtInLiDARDepthCamera,
+        ]
+        let session = AVCaptureDevice.DiscoverySession(
+            deviceTypes: types,
+            mediaType: .video,
+            position: .unspecified)
+        return session.devices
+    }
+
    nonisolated static func clampQuality(_ quality: Double?) -> Double {
        let q = quality ?? 0.9
        return min(1.0, max(0.05, q))
@@ -137,9 +137,11 @@ final class ScreenRecordService: @unchecked Sendable {
        recordQueue: DispatchQueue) -> @Sendable (CMSampleBuffer, RPSampleBufferType, Error?) -> Void
    {
        { sample, type, error in
+            let sampleBox = UncheckedSendableBox(value: sample)
            // ReplayKit can call the capture handler on a background queue.
            // Serialize writes to avoid queue asserts.
            recordQueue.async {
+                let sample = sampleBox.value
                if let error {
                    state.withLock { state in
                        if state.handlerError == nil { state.handlerError = error }
@@ -26,7 +26,8 @@ Sources/Voice/VoiceTab.swift
 Sources/Voice/VoiceWakeManager.swift
 Sources/Voice/VoiceWakePreferences.swift
 ../shared/ClawdbotKit/Sources/ClawdbotChatUI/ChatComposer.swift
-../shared/ClawdbotKit/Sources/ClawdbotChatUI/ChatMarkdownSplitter.swift
+../shared/ClawdbotKit/Sources/ClawdbotChatUI/ChatMarkdownRenderer.swift
+../shared/ClawdbotKit/Sources/ClawdbotChatUI/ChatMarkdownPreprocessor.swift
 ../shared/ClawdbotKit/Sources/ClawdbotChatUI/ChatMessageViews.swift
 ../shared/ClawdbotKit/Sources/ClawdbotChatUI/ChatModels.swift
 ../shared/ClawdbotKit/Sources/ClawdbotChatUI/ChatPayloadDecoding.swift
@@ -2,7 +2,7 @@ name: Clawdbot
 options:
  bundleIdPrefix: com.clawdbot
  deploymentTarget:
-    iOS: "17.0"
+    iOS: "18.0"
  xcodeVersion: "16.0"

 settings:
@@ -1,5 +1,5 @@
 {
-  "originHash" : "9de32b5fc115432dadd84c3ab4d67d2fed22ffaf5675a77033d69ea194ac3862",
+  "originHash" : "7eec77e2b399c480e76fdfc7dc3162652f5c775530e9fc282953de38ef2de79b",
  "pins" : [
    {
      "identity" : "elevenlabskit",
@@ -73,6 +73,15 @@
        "revision" : "8e5e4a8f3617283b556064574651fc0869943c9a"
      }
    },
+    {
+      "identity" : "swift-concurrency-extras",
+      "kind" : "remoteSourceControl",
+      "location" : "https://github.com/pointfreeco/swift-concurrency-extras",
+      "state" : {
+        "revision" : "5a3825302b1a0d744183200915a47b508c828e6f",
+        "version" : "1.3.2"
+      }
+    },
    {
      "identity" : "swift-configuration",
      "kind" : "remoteSourceControl",
@@ -144,6 +153,24 @@
        "revision" : "395a77f0aa927f0ff73941d7ac35f2b46d47c9db",
        "version" : "1.6.3"
      }
+    },
+    {
+      "identity" : "swiftui-math",
+      "kind" : "remoteSourceControl",
+      "location" : "https://github.com/gonzalezreal/swiftui-math",
+      "state" : {
+        "revision" : "0b5c2cfaaec8d6193db206f675048eeb5ce95f71",
+        "version" : "0.1.0"
+      }
+    },
+    {
+      "identity" : "textual",
+      "kind" : "remoteSourceControl",
+      "location" : "https://github.com/gonzalezreal/textual",
+      "state" : {
+        "revision" : "a03c1e103d88de4ea0dd8320ea1611ec0d4b29b3",
+        "version" : "0.2.0"
+      }
    }
  ],
  "version" : 3
@@ -341,6 +341,15 @@ final class AppState {
        return host
    }

+    private static func sanitizeSSHTarget(_ value: String) -> String {
+        let trimmed = value.trimmingCharacters(in: .whitespacesAndNewlines)
+        if trimmed.hasPrefix("ssh ") {
+            return trimmed.replacingOccurrences(of: "ssh ", with: "")
+                .trimmingCharacters(in: .whitespacesAndNewlines)
+        }
+        return trimmed
+    }
+
    private func startConfigWatcher() {
        let configUrl = ClawdbotConfigFile.url()
        self.configWatcher = ConfigFileWatcher(url: configUrl) { [weak self] in
@@ -406,6 +415,7 @@ final class AppState {

        let connectionMode = self.connectionMode
        let remoteTarget = self.remoteTarget
+        let remoteIdentity = self.remoteIdentity
        let desiredMode: String? = switch connectionMode {
        case .local:
            "local"
@@ -435,15 +445,46 @@ final class AppState {
                changed = true
            }

-            if connectionMode == .remote, let host = remoteHost {
+            if connectionMode == .remote {
                var remote = gateway["remote"] as? [String: Any] ?? [:]
-                let existingUrl = (remote["url"] as? String)?.trimmingCharacters(in: .whitespacesAndNewlines) ?? ""
-                let parsedExisting = existingUrl.isEmpty ? nil : URL(string: existingUrl)
-                let scheme = parsedExisting?.scheme?.isEmpty == false ? parsedExisting?.scheme : "ws"
-                let port = parsedExisting?.port ?? 18789
-                let desiredUrl = "\(scheme ?? "ws")://\(host):\(port)"
-                if existingUrl != desiredUrl {
-                    remote["url"] = desiredUrl
+                var remoteChanged = false
+
+                if let host = remoteHost {
+                    let existingUrl = (remote["url"] as? String)?
+                        .trimmingCharacters(in: .whitespacesAndNewlines) ?? ""
+                    let parsedExisting = existingUrl.isEmpty ? nil : URL(string: existingUrl)
+                    let scheme = parsedExisting?.scheme?.isEmpty == false ? parsedExisting?.scheme : "ws"
+                    let port = parsedExisting?.port ?? 18789
+                    let desiredUrl = "\(scheme ?? "ws")://\(host):\(port)"
+                    if existingUrl != desiredUrl {
+                        remote["url"] = desiredUrl
+                        remoteChanged = true
+                    }
+                }
+
+                let sanitizedTarget = Self.sanitizeSSHTarget(remoteTarget)
+                if !sanitizedTarget.isEmpty {
+                    if (remote["sshTarget"] as? String) != sanitizedTarget {
+                        remote["sshTarget"] = sanitizedTarget
+                        remoteChanged = true
+                    }
+                } else if remote["sshTarget"] != nil {
+                    remote.removeValue(forKey: "sshTarget")
+                    remoteChanged = true
+                }
+
+                let trimmedIdentity = remoteIdentity.trimmingCharacters(in: .whitespacesAndNewlines)
+                if !trimmedIdentity.isEmpty {
+                    if (remote["sshIdentity"] as? String) != trimmedIdentity {
+                        remote["sshIdentity"] = trimmedIdentity
+                        remoteChanged = true
+                    }
+                } else if remote["sshIdentity"] != nil {
+                    remote.removeValue(forKey: "sshIdentity")
+                    remoteChanged = true
+                }
+
+                if remoteChanged {
                    gateway["remote"] = remote
                    changed = true
                }
@@ -188,9 +188,15 @@ extension CronJobEditor {
        }
    }

-    func applyDeleteAfterRun(to root: inout [String: Any]) {
-        if self.scheduleKind == .at {
-            root["deleteAfterRun"] = self.deleteAfterRun
+    func applyDeleteAfterRun(
+        to root: inout [String: Any],
+        scheduleKind: ScheduleKind? = nil,
+        deleteAfterRun: Bool? = nil)
+    {
+        let resolvedSchedule = scheduleKind ?? self.scheduleKind
+        let resolvedDelete = deleteAfterRun ?? self.deleteAfterRun
+        if resolvedSchedule == .at {
+            root["deleteAfterRun"] = resolvedDelete
        } else if self.job?.deleteAfterRun != nil {
            root["deleteAfterRun"] = false
        }
@@ -69,7 +69,10 @@ extension GatewayLaunchAgentManager {
    }

    private static func readDaemonLoaded() async -> Bool? {
-        let result = await self.runDaemonCommand(["status", "--json", "--no-probe"], timeout: 15, quiet: true)
+        let result = await self.runDaemonCommandResult(
+            ["status", "--json", "--no-probe"],
+            timeout: 15,
+            quiet: true)
        guard result.success, let payload = result.payload else { return nil }
        guard
            let json = try? JSONSerialization.jsonObject(with: payload) as? [String: Any],
@@ -87,6 +87,14 @@ final class GatewayProcessManager {
            self.status = .stopped
            return
        }
+        // Many surfaces can call `setActive(true)` in quick succession (startup, Canvas, health checks).
+        // Avoid spawning multiple concurrent "start" tasks that can thrash launchd and flap the port.
+        switch self.status {
+        case .starting, .running, .attachedExisting:
+            return
+        case .stopped, .failed:
+            break
+        }
        self.status = .starting
        self.logger.debug("gateway start requested")

@@ -92,10 +92,13 @@ struct GeneralSettings: View {
                                Text(policy.title).tag(policy)
                            }
                        }
-                        .pickerStyle(.segmented)
+                        .labelsHidden()
+                        .pickerStyle(.menu)

                        Text("""
-                        Controls remote command execution on this Mac when it is paired as a node. "Always Ask" prompts on each command; "Always Allow" runs without prompts; "Never" disables `system.run`.
+                        Controls remote command execution on this Mac when it is paired as a node. \
+                        "Always Ask" prompts on each command; "Always Allow" runs without prompts; \
+                        "Never" disables `system.run`.
                        """)
                        .font(.footnote)
                        .foregroundStyle(.tertiary)
@@ -111,7 +114,8 @@ struct GeneralSettings: View {
                            Text("While Using").tag(ClawdbotLocationMode.whileUsing.rawValue)
                            Text("Always").tag(ClawdbotLocationMode.always.rawValue)
                        }
-                        .pickerStyle(.segmented)
+                        .labelsHidden()
+                        .pickerStyle(.menu)

                        Toggle("Precise Location", isOn: self.$locationPreciseEnabled)
                            .disabled(self.locationMode == .off)
@@ -395,7 +395,7 @@ extension InstancesSettings {
            host: "phone",
            ip: "10.0.0.3",
            version: "2.0.0",
-            platform: "iOS 17.2",
+            platform: "iOS 18.0",
            deviceFamily: "iPhone",
            modelIdentifier: nil,
            lastInputSeconds: 35,
@@ -446,7 +446,7 @@ extension InstancesSettings {
        _ = view.platformIcon("watchOS 10")
        _ = view.platformIcon("unknown 1.0")
        _ = view.prettyPlatform("macOS 14.2")
-        _ = view.prettyPlatform("iOS 17")
+        _ = view.prettyPlatform("iOS 18")
        _ = view.prettyPlatform("ipados 17.1")
        _ = view.prettyPlatform("linux")
        _ = view.prettyPlatform("   ")
@@ -460,7 +460,7 @@ actor MacNodeBridgeSession {
        do {
            try await self.send(response)
        } catch {
-            await self.logInvokeSendFailure(error)
+            self.logInvokeSendFailure(error)
        }
    }

@@ -19,14 +19,14 @@ enum MacNodeBridgeTLSStore {
    }

    static func loadFingerprint(stableID: String) -> String? {
-        let key = keyPrefix + stableID
-        let raw = defaults.string(forKey: key)?.trimmingCharacters(in: .whitespacesAndNewlines)
+        let key = self.keyPrefix + stableID
+        let raw = self.defaults.string(forKey: key)?.trimmingCharacters(in: .whitespacesAndNewlines)
        return raw?.isEmpty == false ? raw : nil
    }

    static func saveFingerprint(_ value: String, stableID: String) {
-        let key = keyPrefix + stableID
-        defaults.set(value, forKey: key)
+        let key = self.keyPrefix + stableID
+        self.defaults.set(value, forKey: key)
    }
 }

@@ -40,11 +40,10 @@ func makeMacNodeTLSOptions(_ params: MacNodeBridgeTLSParams?) -> NWProtocolTLS.O
    sec_protocol_options_set_verify_block(
        options.securityProtocolOptions,
        { _, trust, complete in
-            guard let trust else {
-                complete(false)
-                return
-            }
-            if let cert = SecTrustGetCertificateAtIndex(trust, 0) {
+            let trustRef = sec_trust_copy_ref(trust).takeRetainedValue()
+            if let chain = SecTrustCopyCertificateChain(trustRef) as? [SecCertificate],
+               let cert = chain.first
+            {
                let data = SecCertificateCopyData(cert) as Data
                let fingerprint = sha256Hex(data)
                if let expected {
@@ -57,7 +56,7 @@ func makeMacNodeTLSOptions(_ params: MacNodeBridgeTLSParams?) -> NWProtocolTLS.O
                    return
                }
            }
-            let ok = SecTrustEvaluateWithError(trust, nil)
+            let ok = SecTrustEvaluateWithError(trustRef, nil)
            complete(ok)
        },
        DispatchQueue(label: "com.clawdbot.macos.bridge.tls.verify"))
@@ -71,5 +70,5 @@ private func sha256Hex(_ data: Data) -> String {
 }

 private func normalizeMacNodeFingerprint(_ raw: String) -> String {
-    raw.lowercased().filter { $0.isHexDigit }
+    raw.lowercased().filter(\.isHexDigit)
 }
@@ -164,6 +164,7 @@ final class MacNodeModeCoordinator {
        ]

        if SystemRunPolicy.load() != .never {
+            commands.append(ClawdbotSystemCommand.which.rawValue)
            commands.append(ClawdbotSystemCommand.run.rawValue)
        }

@@ -463,7 +464,7 @@ final class MacNodeModeCoordinator {
        }
    }

-    private static func targetFromResult(_ result: NWBrowser.Result) -> BridgeTarget? {
+    private nonisolated static func targetFromResult(_ result: NWBrowser.Result) -> BridgeTarget? {
        let endpoint = result.endpoint
        guard case .service = endpoint else { return nil }
        let stableID = BridgeEndpointID.stableID(endpoint)
@@ -477,7 +478,7 @@ final class MacNodeModeCoordinator {
        return BridgeTarget(endpoint: endpoint, stableID: stableID, tls: tlsParams)
    }

-    private static func resolveDiscoveredTLSParams(
+    private nonisolated static func resolveDiscoveredTLSParams(
        stableID: String,
        tlsEnabled: Bool,
        tlsFingerprintSha256: String?) -> MacNodeBridgeTLSParams?
@@ -503,7 +504,7 @@ final class MacNodeModeCoordinator {
        return nil
    }

-    private static func resolveManualTLSParams(stableID: String) -> MacNodeBridgeTLSParams? {
+    private nonisolated static func resolveManualTLSParams(stableID: String) -> MacNodeBridgeTLSParams? {
        if let stored = MacNodeBridgeTLSStore.loadFingerprint(stableID: stableID) {
            return MacNodeBridgeTLSParams(
                required: true,
@@ -519,12 +520,12 @@ final class MacNodeModeCoordinator {
            storeKey: stableID)
    }

-    private static func txtValue(_ dict: [String: String], key: String) -> String? {
+    private nonisolated static func txtValue(_ dict: [String: String], key: String) -> String? {
        let raw = dict[key]?.trimmingCharacters(in: .whitespacesAndNewlines) ?? ""
        return raw.isEmpty ? nil : raw
    }

-    private static func txtBoolValue(_ dict: [String: String], key: String) -> Bool {
+    private nonisolated static func txtBoolValue(_ dict: [String: String], key: String) -> Bool {
        guard let raw = self.txtValue(dict, key: key)?.lowercased() else { return false }
        return raw == "1" || raw == "true" || raw == "yes"
    }
@@ -55,6 +55,8 @@ actor MacNodeRuntime {
                return try await self.handleScreenRecordInvoke(req)
            case ClawdbotSystemCommand.run.rawValue:
                return try await self.handleSystemRun(req)
+            case ClawdbotSystemCommand.which.rawValue:
+                return try await self.handleSystemWhich(req)
            case ClawdbotSystemCommand.notify.rawValue:
                return try await self.handleSystemNotify(req)
            default:
@@ -494,6 +496,33 @@ actor MacNodeRuntime {
        return BridgeInvokeResponse(id: req.id, ok: true, payloadJSON: payload)
    }

+    private func handleSystemWhich(_ req: BridgeInvokeRequest) async throws -> BridgeInvokeResponse {
+        let params = try Self.decodeParams(ClawdbotSystemWhichParams.self, from: req.paramsJSON)
+        let bins = params.bins
+            .map { $0.trimmingCharacters(in: .whitespacesAndNewlines) }
+            .filter { !$0.isEmpty }
+        guard !bins.isEmpty else {
+            return Self.errorResponse(req, code: .invalidRequest, message: "INVALID_REQUEST: bins required")
+        }
+
+        let searchPaths = CommandResolver.preferredPaths()
+        var matches: [String] = []
+        var paths: [String: String] = [:]
+        for bin in bins {
+            if let path = CommandResolver.findExecutable(named: bin, searchPaths: searchPaths) {
+                matches.append(bin)
+                paths[bin] = path
+            }
+        }
+
+        struct WhichPayload: Encodable {
+            let bins: [String]
+            let paths: [String: String]
+        }
+        let payload = try Self.encodePayload(WhichPayload(bins: matches, paths: paths))
+        return BridgeInvokeResponse(id: req.id, ok: true, payloadJSON: payload)
+    }
+
    private func handleSystemNotify(_ req: BridgeInvokeRequest) async throws -> BridgeInvokeResponse {
        let params = try Self.decodeParams(ClawdbotSystemNotifyParams.self, from: req.paramsJSON)
        let title = params.title.trimmingCharacters(in: .whitespacesAndNewlines)
@@ -586,8 +615,8 @@ actor MacNodeRuntime {
            let key = rawKey.trimmingCharacters(in: .whitespacesAndNewlines)
            guard !key.isEmpty else { continue }
            let upper = key.uppercased()
-            if blockedEnvKeys.contains(upper) { continue }
-            if blockedEnvPrefixes.contains(where: { upper.hasPrefix($0) }) { continue }
+            if self.blockedEnvKeys.contains(upper) { continue }
+            if self.blockedEnvPrefixes.contains(where: { upper.hasPrefix($0) }) { continue }
            merged[key] = value
        }
        return merged
@@ -351,10 +351,11 @@ actor PortGuardian {
            if port == GatewayEnvironment.gatewayPort() { return cmd.contains("ssh") }
            return false
        case .local:
-            if !cmd.contains("clawdbot") { return false }
+            // The gateway daemon may listen as `clawdbot` or as its runtime (`node`, `bun`, etc).
            if full.contains("gateway-daemon") { return true }
            // If args are unavailable, treat a clawdbot listener as expected.
-            return full == cmd
+            if cmd.contains("clawdbot"), full == cmd { return true }
+            return false
        case .unconfigured:
            return false
        }
@@ -10,11 +10,11 @@ enum SystemRunPolicy: String, CaseIterable, Identifiable {
    var title: String {
        switch self {
        case .never:
-            return "Never"
+            "Never"
        case .ask:
-            return "Always Ask"
+            "Always Ask"
        case .always:
-            return "Always Allow"
+            "Always Allow"
        }
    }

@@ -75,13 +75,13 @@ enum SystemRunAllowlist {

    static func contains(_ argv: [String], defaults: UserDefaults = .standard) -> Bool {
        let key = key(for: argv)
-        return load(from: defaults).contains(key)
+        return self.load(from: defaults).contains(key)
    }

    static func add(_ argv: [String], defaults: UserDefaults = .standard) {
        let key = key(for: argv)
        guard !key.isEmpty else { return }
-        var allowlist = load(from: defaults)
+        var allowlist = self.load(from: defaults)
        if allowlist.insert(key).inserted {
            MacNodeConfigFile.setSystemRunAllowlist(Array(allowlist).sorted())
        }
@@ -1,3 +1,4 @@
+import ClawdbotProtocol
 import SwiftUI
 import Testing
@testable import Clawdbot
@@ -38,7 +38,7 @@ struct CronJobEditorSmokeTests {
                thinking: "low",
                timeoutSeconds: 120,
                deliver: true,
-                provider: "whatsapp",
+                channel: "whatsapp",
                to: "+15551234567",
                bestEffortDeliver: true),
            isolation: CronIsolation(postToMainPrefix: "Cron"),
@@ -70,22 +70,16 @@ struct CronJobEditorSmokeTests {
    }

    @Test func cronJobEditorIncludesDeleteAfterRunForAtSchedule() throws {
-        var view = CronJobEditor(
+        let view = CronJobEditor(
            job: nil,
            isSaving: .constant(false),
            error: .constant(nil),
            onCancel: {},
            onSave: { _ in })
-        view.name = "One-shot"
-        view.sessionTarget = .main
-        view.payloadKind = .systemEvent
-        view.systemEventText = "hello"
-        view.scheduleKind = .at
-        view.atDate = Date(timeIntervalSince1970: 1_700_000_000)
-        view.deleteAfterRun = true

-        let payload = try view.buildPayload()
-        let raw = payload["deleteAfterRun"]?.value as? Bool
+        var root: [String: Any] = [:]
+        view.applyDeleteAfterRun(to: &root, scheduleKind: .at, deleteAfterRun: true)
+        let raw = root["deleteAfterRun"] as? Bool
        #expect(raw == true)
    }
 }
@@ -31,7 +31,7 @@ struct CronModelsTests {
            thinking: "low",
            timeoutSeconds: 15,
            deliver: true,
-            provider: "whatsapp",
+            channel: "whatsapp",
            to: "+15551234567",
            bestEffortDeliver: false)
        let data = try JSONEncoder().encode(payload)
@@ -39,7 +39,7 @@ struct InstancesSettingsSmokeTests {
                host: "gateway",
                ip: "10.0.0.4",
                version: "3.0.0",
-                platform: "iOS 17",
+                platform: "iOS 18",
                deviceFamily: nil,
                modelIdentifier: nil,
                lastInputSeconds: nil,
@@ -11,7 +11,8 @@ import Testing
            stateversion: StateVersion(presence: 1, health: 1),
            uptimems: 123,
            configpath: nil,
-            statedir: nil)
+            statedir: nil,
+            sessiondefaults: nil)

        let hello = HelloOk(
            type: "hello",
@@ -21,6 +21,15 @@ struct MacNodeRuntimeTests {
        #expect(response.ok == false)
    }

+    @Test func handleInvokeRejectsEmptySystemWhich() async throws {
+        let runtime = MacNodeRuntime()
+        let params = ClawdbotSystemWhichParams(bins: [])
+        let json = try String(data: JSONEncoder().encode(params), encoding: .utf8)
+        let response = await runtime.handleInvoke(
+            BridgeInvokeRequest(id: "req-2b", command: ClawdbotSystemCommand.which.rawValue, paramsJSON: json))
+        #expect(response.ok == false)
+    }
+
    @Test func handleInvokeRejectsEmptyNotification() async throws {
        let runtime = MacNodeRuntime()
        let params = ClawdbotSystemNotifyParams(title: "", body: "")
@@ -65,6 +74,10 @@ struct MacNodeRuntimeTests {
            {
                CLLocation(latitude: 0, longitude: 0)
            }
+
+            func confirmSystemRun(command: String, cwd: String?) async -> SystemRunDecision {
+                .allowOnce
+            }
        }

        let services = await MainActor.run { FakeMainActorServices() }
@@ -16,13 +16,13 @@ struct OnboardingViewSmokeTests {
    }

    @Test func pageOrderOmitsWorkspaceAndIdentitySteps() {
-        let order = OnboardingView.pageOrder(for: .local, needsBootstrap: false)
+        let order = OnboardingView.pageOrder(for: .local, showOnboardingChat: false)
        #expect(!order.contains(7))
        #expect(order.contains(3))
    }

    @Test func pageOrderOmitsOnboardingChatWhenIdentityKnown() {
-        let order = OnboardingView.pageOrder(for: .local, needsBootstrap: false)
+        let order = OnboardingView.pageOrder(for: .local, showOnboardingChat: false)
        #expect(!order.contains(8))
    }
 }
@@ -49,7 +49,7 @@ struct SettingsViewSmokeTests {
                thinking: "low",
                timeoutSeconds: 30,
                deliver: true,
-                provider: "sms",
+                channel: "sms",
                to: "+15551234567",
                bestEffortDeliver: true),
            isolation: CronIsolation(postToMainPrefix: "[cron] "),
@@ -8,7 +8,7 @@ import Testing
        let wav = makeWav16Mono(sampleRate: 8000, samples: 80)
        defer { _ = TalkAudioPlayer.shared.stop() }

-        _ = try await withTimeout(seconds: 2.0) {
+        _ = try await withTimeout(seconds: 4.0) {
            await TalkAudioPlayer.shared.play(data: wav)
        }

@@ -27,7 +27,7 @@ import Testing
        await Task.yield()
        _ = await TalkAudioPlayer.shared.play(data: wav)

-        _ = try await withTimeout(seconds: 2.0) {
+        _ = try await withTimeout(seconds: 4.0) {
            await first.value
        }
        #expect(true)
@@ -17,6 +17,6 @@ import Testing
        #expect(opts.thinking == "low")
        #expect(opts.deliver == true)
        #expect(opts.to == nil)
-        #expect(opts.provider == .last)
+        #expect(opts.channel == .last)
    }
 }
@@ -5,7 +5,7 @@ import PackageDescription
 let package = Package(
    name: "ClawdbotKit",
    platforms: [
-        .iOS(.v17),
+        .iOS(.v18),
        .macOS(.v15),
    ],
    products: [
@@ -14,6 +14,7 @@ let package = Package(
    ],
    dependencies: [
        .package(url: "https://github.com/steipete/ElevenLabsKit", exact: "0.1.0"),
+        .package(url: "https://github.com/gonzalezreal/textual", exact: "0.2.0"),
    ],
    targets: [
        .target(
@@ -29,7 +30,13 @@ let package = Package(
            ]),
        .target(
            name: "ClawdbotChatUI",
-            dependencies: ["ClawdbotKit"],
+            dependencies: [
+                "ClawdbotKit",
+                .product(
+                    name: "Textual",
+                    package: "textual",
+                    condition: .when(platforms: [.macOS, .iOS])),
+            ],
            swiftSettings: [
                .enableUpcomingFeature("StrictConcurrency"),
            ]),
@@ -0,0 +1,51 @@
+import Foundation
+
+enum ChatMarkdownPreprocessor {
+    struct InlineImage: Identifiable {
+        let id = UUID()
+        let label: String
+        let image: ClawdbotPlatformImage?
+    }
+
+    struct Result {
+        let cleaned: String
+        let images: [InlineImage]
+    }
+
+    static func preprocess(markdown raw: String) -> Result {
+        let pattern = #"!\[([^\]]*)\]\((data:image\/[^;]+;base64,[^)]+)\)"#
+        guard let re = try? NSRegularExpression(pattern: pattern) else {
+            return Result(cleaned: raw, images: [])
+        }
+
+        let ns = raw as NSString
+        let matches = re.matches(in: raw, range: NSRange(location: 0, length: ns.length))
+        if matches.isEmpty { return Result(cleaned: raw, images: []) }
+
+        var images: [InlineImage] = []
+        var cleaned = raw
+
+        for match in matches.reversed() {
+            guard match.numberOfRanges >= 3 else { continue }
+            let label = ns.substring(with: match.range(at: 1))
+            let dataURL = ns.substring(with: match.range(at: 2))
+
+            let image: ClawdbotPlatformImage? = {
+                guard let comma = dataURL.firstIndex(of: ",") else { return nil }
+                let b64 = String(dataURL[dataURL.index(after: comma)...])
+                guard let data = Data(base64Encoded: b64) else { return nil }
+                return ClawdbotPlatformImage(data: data)
+            }()
+            images.append(InlineImage(label: label, image: image))
+
+            let start = cleaned.index(cleaned.startIndex, offsetBy: match.range.location)
+            let end = cleaned.index(start, offsetBy: match.range.length)
+            cleaned.replaceSubrange(start..<end, with: "")
+        }
+
+        let normalized = cleaned
+            .replacingOccurrences(of: "\n\n\n", with: "\n\n")
+            .trimmingCharacters(in: .whitespacesAndNewlines)
+        return Result(cleaned: normalized, images: images.reversed())
+    }
+}
@@ -0,0 +1,90 @@
+import SwiftUI
+import Textual
+
+public enum ChatMarkdownVariant: String, CaseIterable, Sendable {
+    case standard
+    case compact
+}
+
+@MainActor
+struct ChatMarkdownRenderer: View {
+    enum Context {
+        case user
+        case assistant
+    }
+
+    let text: String
+    let context: Context
+    let variant: ChatMarkdownVariant
+    let font: Font
+    let textColor: Color
+
+    var body: some View {
+        let processed = ChatMarkdownPreprocessor.preprocess(markdown: self.text)
+        VStack(alignment: .leading, spacing: 10) {
+            StructuredText(markdown: processed.cleaned)
+                .modifier(ChatMarkdownStyle(
+                    variant: self.variant,
+                    context: self.context,
+                    font: self.font,
+                    textColor: self.textColor))
+
+            if !processed.images.isEmpty {
+                InlineImageList(images: processed.images)
+            }
+        }
+    }
+}
+
+private struct ChatMarkdownStyle: ViewModifier {
+    let variant: ChatMarkdownVariant
+    let context: ChatMarkdownRenderer.Context
+    let font: Font
+    let textColor: Color
+
+    func body(content: Content) -> some View {
+        Group {
+            if self.variant == .compact {
+                content.textual.structuredTextStyle(.default)
+            } else {
+                content.textual.structuredTextStyle(.gitHub)
+            }
+        }
+        .font(self.font)
+        .foregroundStyle(self.textColor)
+        .textual.inlineStyle(self.inlineStyle)
+        .textual.textSelection(.enabled)
+    }
+
+    private var inlineStyle: InlineStyle {
+        let linkColor: Color = self.context == .user ? self.textColor : .accentColor
+        let codeScale: CGFloat = self.variant == .compact ? 0.85 : 0.9
+        return InlineStyle()
+            .code(.monospaced, .fontScale(codeScale))
+            .link(.foregroundColor(linkColor))
+    }
+}
+
+@MainActor
+private struct InlineImageList: View {
+    let images: [ChatMarkdownPreprocessor.InlineImage]
+
+    var body: some View {
+        ForEach(images, id: \.id) { item in
+            if let img = item.image {
+                ClawdbotPlatformImageFactory.image(img)
+                    .resizable()
+                    .scaledToFit()
+                    .frame(maxHeight: 260)
+                    .clipShape(RoundedRectangle(cornerRadius: 12, style: .continuous))
+                    .overlay(
+                        RoundedRectangle(cornerRadius: 12, style: .continuous)
+                            .strokeBorder(Color.white.opacity(0.12), lineWidth: 1))
+            } else {
+                Text(item.label.isEmpty ? "Image" : item.label)
+                    .font(.footnote)
+                    .foregroundStyle(.secondary)
+            }
+        }
+    }
+}
@@ -1,114 +0,0 @@
-import Foundation
-
-enum ChatMarkdownSplitter {
-    struct InlineImage: Identifiable {
-        let id = UUID()
-        let label: String
-        let image: ClawdbotPlatformImage?
-    }
-
-    struct Block: Identifiable {
-        enum Kind: Equatable {
-            case text
-            case code(language: String?)
-        }
-
-        let id = UUID()
-        let kind: Kind
-        let text: String
-    }
-
-    struct SplitResult {
-        let blocks: [Block]
-        let images: [InlineImage]
-    }
-
-    static func split(markdown raw: String) -> SplitResult {
-        let extracted = self.extractInlineImages(from: raw)
-        let blocks = self.splitCodeBlocks(from: extracted.cleaned)
-        return SplitResult(blocks: blocks, images: extracted.images)
-    }
-
-    private static func splitCodeBlocks(from raw: String) -> [Block] {
-        var blocks: [Block] = []
-        var buffer: [String] = []
-        var inCode = false
-        var codeLang: String?
-        var codeLines: [String] = []
-
-        for line in raw.split(separator: "\n", omittingEmptySubsequences: false).map(String.init) {
-            if line.hasPrefix("```") {
-                if inCode {
-                    blocks.append(Block(kind: .code(language: codeLang), text: codeLines.joined(separator: "\n")))
-                    codeLines.removeAll(keepingCapacity: true)
-                    inCode = false
-                    codeLang = nil
-                } else {
-                    let text = buffer.joined(separator: "\n").trimmingCharacters(in: .whitespacesAndNewlines)
-                    if !text.isEmpty {
-                        blocks.append(Block(kind: .text, text: text))
-                    }
-                    buffer.removeAll(keepingCapacity: true)
-                    inCode = true
-                    codeLang = line.dropFirst(3).trimmingCharacters(in: .whitespacesAndNewlines)
-                    if codeLang?.isEmpty == true { codeLang = nil }
-                }
-                continue
-            }
-
-            if inCode {
-                codeLines.append(line)
-            } else {
-                buffer.append(line)
-            }
-        }
-
-        if inCode {
-            blocks.append(Block(kind: .code(language: codeLang), text: codeLines.joined(separator: "\n")))
-        } else {
-            let text = buffer.joined(separator: "\n").trimmingCharacters(in: .whitespacesAndNewlines)
-            if !text.isEmpty {
-                blocks.append(Block(kind: .text, text: text))
-            }
-        }
-
-        return blocks.isEmpty ? [Block(kind: .text, text: raw)] : blocks
-    }
-
-    private static func extractInlineImages(from raw: String) -> (cleaned: String, images: [InlineImage]) {
-        let pattern = #"!\[([^\]]*)\]\((data:image\/[^;]+;base64,[^)]+)\)"#
-        guard let re = try? NSRegularExpression(pattern: pattern) else {
-            return (raw, [])
-        }
-
-        let ns = raw as NSString
-        let matches = re.matches(in: raw, range: NSRange(location: 0, length: ns.length))
-        if matches.isEmpty { return (raw, []) }
-
-        var images: [InlineImage] = []
-        var cleaned = raw
-
-        for match in matches.reversed() {
-            guard match.numberOfRanges >= 3 else { continue }
-            let label = ns.substring(with: match.range(at: 1))
-            let dataURL = ns.substring(with: match.range(at: 2))
-
-            let image: ClawdbotPlatformImage? = {
-                guard let comma = dataURL.firstIndex(of: ",") else { return nil }
-                let b64 = String(dataURL[dataURL.index(after: comma)...])
-                guard let data = Data(base64Encoded: b64) else { return nil }
-                return ClawdbotPlatformImage(data: data)
-            }()
-            images.append(InlineImage(label: label, image: image))
-
-            let start = cleaned.index(cleaned.startIndex, offsetBy: match.range.location)
-            let end = cleaned.index(start, offsetBy: match.range.length)
-            cleaned.replaceSubrange(start..<end, with: "")
-        }
-
-        let normalized = cleaned
-            .replacingOccurrences(of: "\n\n\n", with: "\n\n")
-            .trimmingCharacters(in: .whitespacesAndNewlines)
-        return (normalized, images.reversed())
-    }
-}
@@ -137,10 +137,16 @@ private struct ChatBubbleShape: InsettableShape {
 struct ChatMessageBubble: View {
    let message: ClawdbotChatMessage
    let style: ClawdbotChatView.Style
+    let markdownVariant: ChatMarkdownVariant
    let userAccent: Color?

    var body: some View {
-        ChatMessageBody(message: self.message, isUser: self.isUser, style: self.style, userAccent: self.userAccent)
+        ChatMessageBody(
+            message: self.message,
+            isUser: self.isUser,
+            style: self.style,
+            markdownVariant: self.markdownVariant,
+            userAccent: self.userAccent)
            .frame(maxWidth: ChatUIConstants.bubbleMaxWidth, alignment: self.isUser ? .trailing : .leading)
            .frame(maxWidth: .infinity, alignment: self.isUser ? .trailing : .leading)
            .padding(.horizontal, 2)
@@ -154,6 +160,7 @@ private struct ChatMessageBody: View {
    let message: ClawdbotChatMessage
    let isUser: Bool
    let style: ClawdbotChatView.Style
+    let markdownVariant: ChatMarkdownVariant
    let userAccent: Color?

    var body: some View {
@@ -169,39 +176,14 @@ private struct ChatMessageBody: View {
                        isUser: self.isUser)
                }
            } else if self.isUser {
-                let split = ChatMarkdownSplitter.split(markdown: text)
-                ForEach(split.blocks) { block in
-                    switch block.kind {
-                    case .text:
-                        MarkdownTextView(text: block.text, textColor: textColor, font: .system(size: 14))
-                    case let .code(language):
-                        CodeBlockView(code: block.text, language: language, isUser: self.isUser)
-                    }
-                }
-
-                if !split.images.isEmpty {
-                    ForEach(
-                        split.images,
-                        id: \ChatMarkdownSplitter.InlineImage.id)
-                    { (item: ChatMarkdownSplitter.InlineImage) in
-                        if let img = item.image {
-                            ClawdbotPlatformImageFactory.image(img)
-                                .resizable()
-                                .scaledToFit()
-                                .frame(maxHeight: 260)
-                                .clipShape(RoundedRectangle(cornerRadius: 12, style: .continuous))
-                                .overlay(
-                                    RoundedRectangle(cornerRadius: 12, style: .continuous)
-                                        .strokeBorder(Color.white.opacity(0.12), lineWidth: 1))
-                        } else {
-                            Text(item.label.isEmpty ? "Image" : item.label)
-                                .font(.footnote)
-                                .foregroundStyle(.secondary)
-                        }
-                    }
-                }
+                ChatMarkdownRenderer(
+                    text: text,
+                    context: .user,
+                    variant: self.markdownVariant,
+                    font: .system(size: 14),
+                    textColor: textColor)
            } else {
-                ChatAssistantTextBody(text: text)
+                ChatAssistantTextBody(text: text, markdownVariant: self.markdownVariant)
            }

            if !self.inlineAttachments.isEmpty {
@@ -505,10 +487,11 @@ extension ChatTypingIndicatorBubble: @MainActor Equatable {
@MainActor
 struct ChatStreamingAssistantBubble: View {
    let text: String
+    let markdownVariant: ChatMarkdownVariant

    var body: some View {
        VStack(alignment: .leading, spacing: 10) {
-            ChatAssistantTextBody(text: self.text)
+            ChatAssistantTextBody(text: self.text, markdownVariant: self.markdownVariant)
        }
        .padding(12)
        .background(
@@ -612,114 +595,22 @@ private struct TypingDots: View {
    }
 }

-@MainActor
-private struct MarkdownTextView: View {
-    let text: String
-    let textColor: Color
-    let font: Font
-
-    var body: some View {
-        let normalized = self.text.replacingOccurrences(
-            of: "(?<!\\n)\\n(?!\\n)",
-            with: " ",
-            options: .regularExpression)
-        let options = AttributedString.MarkdownParsingOptions(
-            interpretedSyntax: .inlineOnlyPreservingWhitespace)
-        if let attributed = try? AttributedString(markdown: normalized, options: options) {
-            Text(attributed)
-                .font(self.font)
-                .foregroundStyle(self.textColor)
-        } else {
-            Text(normalized)
-                .font(self.font)
-                .foregroundStyle(self.textColor)
-        }
-    }
-}
-
-@MainActor
 private struct ChatAssistantTextBody: View {
    let text: String
+    let markdownVariant: ChatMarkdownVariant

    var body: some View {
        let segments = AssistantTextParser.segments(from: self.text)
        VStack(alignment: .leading, spacing: 10) {
            ForEach(segments) { segment in
                let font = segment.kind == .thinking ? Font.system(size: 14).italic() : Font.system(size: 14)
-                ChatMarkdownBody(text: segment.text, textColor: ClawdbotChatTheme.assistantText, font: font)
+                ChatMarkdownRenderer(
+                    text: segment.text,
+                    context: .assistant,
+                    variant: self.markdownVariant,
+                    font: font,
+                    textColor: ClawdbotChatTheme.assistantText)
            }
        }
    }
 }
-
-@MainActor
-private struct ChatMarkdownBody: View {
-    let text: String
-    let textColor: Color
-    let font: Font
-
-    var body: some View {
-        let split = ChatMarkdownSplitter.split(markdown: self.text)
-        VStack(alignment: .leading, spacing: 10) {
-            ForEach(split.blocks) { block in
-                switch block.kind {
-                case .text:
-                    MarkdownTextView(text: block.text, textColor: self.textColor, font: self.font)
-                case let .code(language):
-                    CodeBlockView(code: block.text, language: language, isUser: false)
-                }
-            }
-
-            if !split.images.isEmpty {
-                ForEach(
-                    split.images,
-                    id: \ChatMarkdownSplitter.InlineImage.id)
-                { (item: ChatMarkdownSplitter.InlineImage) in
-                    if let img = item.image {
-                        ClawdbotPlatformImageFactory.image(img)
-                            .resizable()
-                            .scaledToFit()
-                            .frame(maxHeight: 260)
-                            .clipShape(RoundedRectangle(cornerRadius: 12, style: .continuous))
-                            .overlay(
-                                RoundedRectangle(cornerRadius: 12, style: .continuous)
-                                    .strokeBorder(Color.white.opacity(0.12), lineWidth: 1))
-                    } else {
-                        Text(item.label.isEmpty ? "Image" : item.label)
-                            .font(.footnote)
-                            .foregroundStyle(.secondary)
-                    }
-                }
-            }
-        }
-        .textSelection(.enabled)
-    }
-}
-
-@MainActor
-private struct CodeBlockView: View {
-    let code: String
-    let language: String?
-    let isUser: Bool
-
-    var body: some View {
-        VStack(alignment: .leading, spacing: 8) {
-            if let language, !language.isEmpty {
-                Text(language)
-                    .font(.caption2.monospaced())
-                    .foregroundStyle(.secondary)
-            }
-            Text(self.code)
-                .font(.system(size: 13, weight: .regular, design: .monospaced))
-                .foregroundStyle(self.isUser ? .white : .primary)
-                .textSelection(.enabled)
-        }
-        .padding(12)
-        .frame(maxWidth: .infinity, alignment: .leading)
-        .background(self.isUser ? Color.white.opacity(0.16) : Color.black.opacity(0.06))
-        .overlay(
-            RoundedRectangle(cornerRadius: 12, style: .continuous)
-                .strokeBorder(Color.black.opacity(0.08), lineWidth: 1))
-        .clipShape(RoundedRectangle(cornerRadius: 12, style: .continuous))
-    }
-}
@@ -14,6 +14,7 @@ public struct ClawdbotChatView: View {
    @State private var hasPerformedInitialScroll = false
    private let showsSessionSwitcher: Bool
    private let style: Style
+    private let markdownVariant: ChatMarkdownVariant
    private let userAccent: Color?

    private enum Layout {
@@ -42,11 +43,13 @@ public struct ClawdbotChatView: View {
        viewModel: ClawdbotChatViewModel,
        showsSessionSwitcher: Bool = false,
        style: Style = .standard,
+        markdownVariant: ChatMarkdownVariant = .standard,
        userAccent: Color? = nil)
    {
        self._viewModel = State(initialValue: viewModel)
        self.showsSessionSwitcher = showsSessionSwitcher
        self.style = style
+        self.markdownVariant = markdownVariant
        self.userAccent = userAccent
    }

@@ -151,7 +154,11 @@ public struct ClawdbotChatView: View {
    @ViewBuilder
    private var messageListRows: some View {
        ForEach(self.visibleMessages) { msg in
-            ChatMessageBubble(message: msg, style: self.style, userAccent: self.userAccent)
+            ChatMessageBubble(
+                message: msg,
+                style: self.style,
+                markdownVariant: self.markdownVariant,
+                userAccent: self.userAccent)
                .frame(
                    maxWidth: .infinity,
                    alignment: msg.role.lowercased() == "user" ? .trailing : .leading)
@@ -172,7 +179,7 @@ public struct ClawdbotChatView: View {
        }

        if let text = self.viewModel.streamingAssistantText, AssistantTextParser.hasVisibleContent(in: text) {
-            ChatStreamingAssistantBubble(text: text)
+            ChatStreamingAssistantBubble(text: text, markdownVariant: self.markdownVariant)
                .frame(maxWidth: .infinity, alignment: .leading)
        }
    }
@@ -2,6 +2,7 @@ import Foundation

 public enum ClawdbotSystemCommand: String, Codable, Sendable {
    case run = "system.run"
+    case which = "system.which"
    case notify = "system.notify"
 }

@@ -39,6 +40,14 @@ public struct ClawdbotSystemRunParams: Codable, Sendable, Equatable {
    }
 }

+public struct ClawdbotSystemWhichParams: Codable, Sendable, Equatable {
+    public var bins: [String]
+
+    public init(bins: [String]) {
+        self.bins = bins
+    }
+}
+
 public struct ClawdbotSystemNotifyParams: Codable, Sendable, Equatable {
    public var title: String
    public var body: String
@@ -0,0 +1,20 @@
+import Testing
+@testable import ClawdbotChatUI
+
+@Suite("ChatMarkdownPreprocessor")
+struct ChatMarkdownPreprocessorTests {
+    @Test func extractsDataURLImages() {
+        let base64 = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVQIHWP4////GQAJ+wP/2hN8NwAAAABJRU5ErkJggg=="
+        let markdown = """
+        Hello
+
+        ![Pixel](data:image/png;base64,\(base64))
+        """
+
+        let result = ChatMarkdownPreprocessor.preprocess(markdown: markdown)
+
+        #expect(result.cleaned == "Hello")
+        #expect(result.images.count == 1)
+        #expect(result.images.first?.image != nil)
+    }
+}
@@ -227,6 +227,8 @@ Outbound Discord API calls retry on rate limits (429) using Discord `retry_after
      actions: {
        reactions: true,
        stickers: true,
+        emojiUploads: true,
+        stickerUploads: true,
        polls: true,
        permissions: true,
        messages: true,
@@ -237,6 +239,7 @@ Outbound Discord API calls retry on rate limits (429) using Discord `retry_after
        roleInfo: true,
        roles: false,
        channelInfo: true,
+        channels: true,
        voiceStatus: true,
        events: true,
        moderation: false
@@ -304,8 +307,9 @@ ack reaction after the bot replies.
 - `retry`: retry policy for outbound Discord API calls (attempts, minDelayMs, maxDelayMs, jitter).
 - `actions`: per-action tool gates; omit to allow all (set `false` to disable).
  - `reactions` (covers react + read reactions)
-  - `stickers`, `polls`, `permissions`, `messages`, `threads`, `pins`, `search`
+  - `stickers`, `emojiUploads`, `stickerUploads`, `polls`, `permissions`, `messages`, `threads`, `pins`, `search`
  - `memberInfo`, `roleInfo`, `channelInfo`, `voiceStatus`, `events`
+  - `channels` (create/edit/delete channels + categories + permissions)
  - `roles` (role add/remove, default `false`)
  - `moderation` (timeout/kick/ban, default `false`)

@@ -321,6 +325,8 @@ Reaction notifications use `guilds.<id>.reactionNotifications`:
 | --- | --- | --- |
 | reactions | enabled | React + list reactions + emojiList |
 | stickers | enabled | Send stickers |
+| emojiUploads | enabled | Upload emojis |
+| stickerUploads | enabled | Upload stickers |
 | polls | enabled | Create polls |
 | permissions | enabled | Channel permission snapshot |
 | messages | enabled | Read/send/edit/delete |
@@ -330,6 +336,7 @@ Reaction notifications use `guilds.<id>.reactionNotifications`:
 | memberInfo | enabled | Member info |
 | roleInfo | enabled | Role list |
 | channelInfo | enabled | Channel info + list |
+| channels | enabled | Channel/category management |
 | voiceStatus | enabled | Voice state lookup |
 | events | enabled | List/create scheduled events |
 | roles | disabled | Role add/remove |
@@ -101,6 +101,29 @@ group messages, so use admin if you need full visibility.
 - Raw HTML from models is escaped to avoid Telegram parse errors.
 - If Telegram rejects the HTML payload, Clawdbot retries the same message as plain text.

+## Commands (native + custom)
+Clawdbot registers native commands (like `/status`, `/reset`, `/model`) with Telegram’s bot menu on startup.
+You can add custom commands to the menu via config:
+
+```json5
+{
+  channels: {
+    telegram: {
+      customCommands: [
+        { command: "backup", description: "Git backup" },
+        { command: "generate", description: "Create an image" }
+      ]
+    }
+  }
+}
+```
+
+Notes:
+- Custom commands are **menu entries only**; Clawdbot does not implement them unless you handle them elsewhere.
+- Command names are normalized (leading `/` stripped, lowercased) and must match `a-z`, `0-9`, `_` (1–32 chars).
+- Custom commands **cannot override native commands**. Conflicts are ignored and logged.
+- If `commands.native` is disabled, only custom commands are registered (or cleared if none).
+
 ## Limits
 - Outbound text is chunked to `channels.telegram.textChunkLimit` (default 4000).
 - Media downloads/uploads are capped by `channels.telegram.mediaMaxMb` (default 5).
@@ -0,0 +1,41 @@
+---
+summary: "CLI reference for `clawdbot config` (get/set/unset config values)"
+read_when:
+  - You want to read or edit config non-interactively
+---
+
+# `clawdbot config`
+
+Config helpers: get/set/unset values by path. Run without a subcommand to open
+the configure wizard (same as `clawdbot configure`).
+
+## Examples
+
+```bash
+clawdbot config get browser.executablePath
+clawdbot config set browser.executablePath "/usr/bin/google-chrome"
+clawdbot config set agents.defaults.heartbeat.every "2h"
+clawdbot config unset tools.web.search.apiKey
+```
+
+## Paths
+
+Paths use dot or bracket notation:
+
+```bash
+clawdbot config get agents.defaults.workspace
+clawdbot config get agents.list[0].id
+```
+
+## Values
+
+Values are parsed as JSON5 when possible; otherwise they are treated as strings.
+Use `--json` to require JSON5 parsing.
+
+```bash
+clawdbot config set agents.defaults.heartbeat.every "0m"
+clawdbot config set gateway.port 19001 --json
+clawdbot config set channels.whatsapp.groups '["*"]' --json
+```
+
+Restart the gateway after edits.
@@ -1,15 +1,19 @@
 ---
-summary: "CLI reference for `clawdbot configure` / `clawdbot config` (interactive configuration prompts)"
+summary: "CLI reference for `clawdbot configure` (interactive configuration prompts)"
 read_when:
  - You want to tweak credentials, devices, or agent defaults interactively
 ---

-# `clawdbot configure` (alias: `config`)
+# `clawdbot configure`

 Interactive prompt to set up credentials, devices, and agent defaults.

+Tip: `clawdbot config` without a subcommand opens the same wizard. Use
+`clawdbot config get|set|unset` for non-interactive edits.
+
 Related:
 - Gateway configuration reference: [Configuration](/gateway/configuration)
+- Config CLI: [Config](/cli/config)

 Notes:
 - Choosing where the Gateway runs always updates `gateway.mode`. You can select "Continue" without other sections if that is all you need.
@@ -13,7 +13,8 @@ This page describes the current CLI behavior. If commands change, update this do

 - [`setup`](/cli/setup)
 - [`onboard`](/cli/onboard)
- [`configure`](/cli/configure) (alias: `config`)
+- [`configure`](/cli/configure)
+- [`config`](/cli/config)
 - [`doctor`](/cli/doctor)
 - [`dashboard`](/cli/dashboard)
 - [`reset`](/cli/reset)
@@ -83,7 +84,11 @@ Palette source of truth: `src/terminal/palette.ts` (aka “lobster seam”).
 clawdbot [--dev] [--profile <name>] <command>
  setup
  onboard
-  configure (alias: config)
+  configure
+  config
+    get
+    set
+    unset
  doctor
  security
    audit
@@ -310,9 +315,18 @@ Options:
 - `--node-manager <npm|pnpm|bun>` (pnpm recommended; bun not recommended for Gateway runtime)
 - `--json`

-### `configure` / `config`
+### `configure`
 Interactive configuration wizard (models, channels, skills, gateway).

+### `config`
+Non-interactive config helpers (get/set/unset). Running `clawdbot config` with no
+subcommand launches the wizard.
+
+Subcommands:
+- `config get <path>`: print a config value (dot/bracket path).
+- `config set <path> <value>`: set a value (JSON5 or raw string).
+- `config unset <path>`: remove a value.
+
 ### `doctor`
 Health checks + quick fixes (config + gateway + legacy services).

@@ -21,6 +21,8 @@ clawdbot plugins info <id>
 clawdbot plugins enable <id>
 clawdbot plugins disable <id>
 clawdbot plugins doctor
+clawdbot plugins update <id>
+clawdbot plugins update --all
 ```

 ### Install
@@ -31,3 +33,12 @@ clawdbot plugins install <npm-spec>

 Security note: treat plugin installs like running code. Prefer pinned versions.

+### Update
+
+```bash
+clawdbot plugins update <id>
+clawdbot plugins update --all
+clawdbot plugins update <id> --dry-run
+```
+
+Updates only apply to plugins installed from npm (tracked in `plugins.installs`).
@@ -871,6 +871,7 @@ Notes:
 - `commands.text: false` disables parsing chat messages for commands.
 - `commands.native: "auto"` (default) turns on native commands for Discord/Telegram and leaves Slack off; unsupported channels stay text-only.
 - Set `commands.native: true|false` to force all, or override per channel with `channels.discord.commands.native`, `channels.telegram.commands.native`, `channels.slack.commands.native` (bool or `"auto"`). `false` clears previously registered commands on Discord/Telegram at startup; Slack commands are managed in the Slack app.
+- `channels.telegram.customCommands` adds extra Telegram bot menu entries. Names are normalized; conflicts with native commands are ignored.
 - `commands.bash: true` enables `! <cmd>` to run host shell commands (`/bash <cmd>` also works as an alias). Requires `tools.elevated.enabled` and allowlisting the sender in `tools.elevated.allowFrom.<channel>`.
 - `commands.bashForegroundMs` controls how long bash waits before backgrounding. While a bash job is running, new `! <cmd>` requests are rejected (one at a time).
 - `commands.config: true` enables `/config` (reads/writes `clawdbot.json`).
@@ -929,6 +930,10 @@ Set `channels.telegram.configWrites: false` to block Telegram-initiated config w
          }
        }
      },
+      customCommands: [
+        { command: "backup", description: "Git backup" },
+        { command: "generate", description: "Create an image" }
+      ],
      historyLimit: 50,                     // include last N group messages as context (0 disables)
      replyToMode: "first",                 // off | first | all
      streamMode: "partial",               // off | partial | block (draft streaming; separate from block streaming)
@@ -161,6 +161,15 @@ Even with strong system prompts, **prompt injection is not solved**. What helps
 - Run sensitive tool execution in a sandbox; keep secrets out of the agent’s reachable filesystem.
 - **Model choice matters:** older/legacy models can be less robust against prompt injection and tool misuse. Prefer modern, instruction-hardened models for any bot with tools. We recommend Anthropic Opus 4.5 because it’s quite good at recognizing prompt injections (see [“A step forward on safety”](https://www.anthropic.com/news/claude-opus-4-5)).

+### Model strength (security note)
+
+Prompt injection resistance is **not** uniform across model tiers. Smaller/cheaper models are generally more susceptible to tool misuse and instruction hijacking, especially under adversarial prompts.
+
+Recommendations:
+- **Use the latest generation, best-tier model** for any bot that can run tools or touch files/networks.
+- **Avoid weaker tiers** (for example, Sonnet or Haiku) for tool-enabled agents or untrusted inboxes.
+- If you must use a smaller model, **reduce blast radius** (read-only tools, strong sandboxing, minimal filesystem access, strict allowlists).
+
 ## Reasoning & verbose output in groups

 `/reasoning` and `/verbose` can expose internal reasoning or tool output that
@@ -169,6 +169,20 @@ The agent was interrupted mid-response.

 **Fix:** Just send another message. The session continues.

+### "Agent failed before reply: Unknown model: anthropic/claude-haiku-3-5"
+
+Clawdbot intentionally rejects **older/insecure models** (especially those more
+vulnerable to prompt injection). If you see this error, the model name is no
+longer supported.
+
+**Fix:**
+- Pick a **latest** model for the provider and update your config or model alias.
+- If you’re unsure which models are available, run `clawdbot models list` or
+  `clawdbot models scan` and choose a supported one.
+- Check gateway logs for the detailed failure reason.
+
+See also: [Models CLI](/cli/models) and [Model providers](/concepts/model-providers).
+
 ### Messages Not Triggering

 **Check 1:** Is the sender allowlisted?
@@ -306,6 +320,139 @@ Clawdbot keeps conversation history in memory.
 }
 ```

+## Common troubleshooting
+
+### “All models failed” — what should I check first?
+
+- **Credentials** present for the provider(s) being tried (auth profiles + env vars).
+- **Model routing**: confirm `agents.defaults.model.primary` and fallbacks are models you can access.
+- **Gateway logs** in `/tmp/clawdbot/…` for the exact provider error.
+- **Model status**: use `/model status` (chat) or `clawdbot models status` (CLI).
+
+### I’m running on my personal WhatsApp number — why is self-chat weird?
+
+Enable self-chat mode and allowlist your own number:
+
+```json5
+{
+  channels: {
+    whatsapp: {
+      selfChatMode: true,
+      dmPolicy: "allowlist",
+      allowFrom: ["+15555550123"]
+    }
+  }
+}
+```
+
+See [WhatsApp setup](/channels/whatsapp).
+
+### WhatsApp logged me out. How do I re‑auth?
+
+Run the login command again and scan the QR code:
+
+```bash
+clawdbot channels login
+```
+
+### Build errors on `main` — what’s the standard fix path?
+
+1) `git pull origin main && pnpm install`
+2) `pnpm clawdbot doctor`
+3) Check GitHub issues or Discord
+4) Temporary workaround: check out an older commit
+
+### npm install fails (allow-build-scripts / missing tar or yargs). What now?
+
+If you’re running from source, use the repo’s package manager: **pnpm** (preferred).
+The repo declares `packageManager: "pnpm@…"`, and pnpm patches are tracked in `pnpm.patchedDependencies`.
+
+Typical recovery:
+```bash
+git status   # ensure you’re in the repo root
+pnpm install
+pnpm build
+pnpm clawdbot doctor
+clawdbot daemon restart
+```
+
+Why: pnpm is the configured package manager for this repo, and the dependency
+patching workflow relies on it.
+
+### How do I switch between git installs and npm installs?
+
+Use the **website installer** and select the install method with a flag. It
+upgrades in place and rewrites the gateway service to point at the new install.
+
+Switch **to git install**:
+```bash
+curl -fsSL https://clawd.bot/install.sh | bash -s -- --install-method git --no-onboard
+```
+
+Switch **to npm global**:
+```bash
+curl -fsSL https://clawd.bot/install.sh | bash
+```
+
+Notes:
+- The git flow only rebases if the repo is clean. Commit or stash changes first.
+- After switching, run:
+  ```bash
+  clawdbot doctor
+  clawdbot daemon restart
+  ```
+
+### Telegram block streaming isn’t splitting text between tool calls. Why?
+
+Block streaming only sends **completed text blocks**. Common reasons you see a single message:
+- `agents.defaults.blockStreamingDefault` is still `"off"`.
+- `channels.telegram.blockStreaming` is set to `false`.
+- `channels.telegram.streamMode` is `partial` or `block` **and draft streaming is active**
+  (private chat + topics). Draft streaming disables block streaming in that case.
+- Your `minChars` / coalesce settings are too high, so chunks get merged.
+- The model emits one large text block (no mid‑reply flush points).
+
+Fix checklist:
+1) Put block streaming settings under `agents.defaults`, not the root.
+2) Set `channels.telegram.streamMode: "off"` if you want real multi‑message block replies.
+3) Use smaller chunk/coalesce thresholds while debugging.
+
+See [Streaming](/concepts/streaming).
+
+### Discord doesn’t reply in my server even with `requireMention: false`. Why?
+
+`requireMention` only controls mention‑gating **after** the channel passes allowlists.
+By default `channels.discord.groupPolicy` is **allowlist**, so guilds must be explicitly enabled.
+If you set `channels.discord.guilds.<guildId>.channels`, only the listed channels are allowed; omit it to allow all channels in the guild.
+
+Fix checklist:
+1) Set `channels.discord.groupPolicy: "open"` **or** add a guild allowlist entry (and optionally a channel allowlist).
+2) Use **numeric channel IDs** in `channels.discord.guilds.<guildId>.channels`.
+3) Put `requireMention: false` **under** `channels.discord.guilds` (global or per‑channel).
+   Top‑level `channels.discord.requireMention` is not a supported key.
+4) Ensure the bot has **Message Content Intent** and channel permissions.
+5) Run `clawdbot channels status --probe` for audit hints.
+
+Docs: [Discord](/channels/discord), [Channels troubleshooting](/channels/troubleshooting).
+
+### Cloud Code Assist API error: invalid tool schema (400). What now?
+
+This is almost always a **tool schema compatibility** issue. The Cloud Code Assist
+endpoint accepts a strict subset of JSON Schema. Clawdbot scrubs/normalizes tool
+schemas in current `main`, but the fix is not in the last release yet (as of
+January 13, 2026).
+
+Fix checklist:
+1) **Update Clawdbot**:
+   - If you can run from source, pull `main` and restart the gateway.
+   - Otherwise, wait for the next release that includes the schema scrubber.
+2) Avoid unsupported keywords like `anyOf/oneOf/allOf`, `patternProperties`,
+   `additionalProperties`, `minLength`, `maxLength`, `format`, etc.
+3) If you define custom tools, keep the top‑level schema as `type: "object"` with
+   `properties` and simple enums.
+
+See [Tools](/tools) and [TypeBox schemas](/concepts/typebox).
+
 ## macOS Specific Issues

 ### App Crashes when Granting Permissions (Speech/Mic)
@@ -160,11 +160,15 @@ clawdbot plugins install <path>              # add a local file/dir to plugins.l
 clawdbot plugins install ./extensions/voice-call # relative path ok
 clawdbot plugins install ./plugin.tgz        # install from a local tarball
 clawdbot plugins install @clawdbot/voice-call # install from npm
+clawdbot plugins update <id>
+clawdbot plugins update --all
 clawdbot plugins enable <id>
 clawdbot plugins disable <id>
 clawdbot plugins doctor
 ```

+`plugins update` only works for npm installs tracked under `plugins.installs`.
+
 Plugins may also register their own top‑level commands (example: `clawdbot voicecall`).

 ## Plugin API (overview)
@@ -70,11 +70,14 @@ Quick answers plus deeper troubleshooting for real-world setups (local dev, VPS,
  - [I set `COPILOT_GITHUB_TOKEN`, but models status shows “Shell env: off.” Why?](#i-set-copilot_github_token-but-models-status-shows-shell-env-off-why)
 - [Sessions & multiple chats](#sessions-multiple-chats)
  - [How do I start a fresh conversation?](#how-do-i-start-a-fresh-conversation)
-  - [How do I completely reset Clawdbot (but keep it installed)?](#how-do-i-completely-reset-clawdbot-but-keep-it-installed)
+  - [Do sessions reset automatically if I never send `/new`?](#do-sessions-reset-automatically-if-i-never-send-new)
+  - [How do I completely reset Clawdbot but keep it installed?](#how-do-i-completely-reset-clawdbot-but-keep-it-installed)
  - [I’m getting “context too large” errors — how do I reset or compact?](#im-getting-context-too-large-errors-how-do-i-reset-or-compact)
+  - [Why am I getting heartbeat messages every 30 minutes?](#why-am-i-getting-heartbeat-messages-every-30-minutes)
  - [Do I need to add a “bot account” to a WhatsApp group?](#do-i-need-to-add-a-bot-account-to-a-whatsapp-group)
  - [Why doesn’t Clawdbot reply in a group?](#why-doesnt-clawdbot-reply-in-a-group)
  - [Do groups/threads share context with DMs?](#do-groupsthreads-share-context-with-dms)
+  - [How many workspaces and agents can I create?](#how-many-workspaces-and-agents-can-i-create)
 - [Models: defaults, selection, aliases, switching](#models-defaults-selection-aliases-switching)
  - [What is the “default model”?](#what-is-the-default-model)
  - [How do I switch models on the fly (without restarting)?](#how-do-i-switch-models-on-the-fly-without-restarting)
@@ -107,25 +110,17 @@ Quick answers plus deeper troubleshooting for real-world setups (local dev, VPS,
 - [Logging and debugging](#logging-and-debugging)
  - [Where are logs?](#where-are-logs)
  - [How do I start/stop/restart the Gateway daemon?](#how-do-i-startstoprestart-the-gateway-daemon)
+  - [ELI5: `clawdbot daemon restart` vs `clawdbot gateway`](#eli5-clawdbot-daemon-restart-vs-clawdbot-gateway)
  - [What’s the fastest way to get more details when something fails?](#whats-the-fastest-way-to-get-more-details-when-something-fails)
 - [Media & attachments](#media-attachments)
  - [My skill generated an image/PDF, but nothing was sent](#my-skill-generated-an-imagepdf-but-nothing-was-sent)
 - [Security and access control](#security-and-access-control)
  - [Is it safe to expose Clawdbot to inbound DMs?](#is-it-safe-to-expose-clawdbot-to-inbound-dms)
+  - [I ran `/start` in Telegram but didn’t get a pairing code](#i-ran-start-in-telegram-but-didnt-get-a-pairing-code)
  - [WhatsApp: will it message my contacts? How does pairing work?](#whatsapp-will-it-message-my-contacts-how-does-pairing-work)
 - [Chat commands, aborting tasks, and “it won’t stop”](#chat-commands-aborting-tasks-and-it-wont-stop)
  - [How do I stop/cancel a running task?](#how-do-i-stopcancel-a-running-task)
  - [Why does it feel like the bot “ignores” rapid‑fire messages?](#why-does-it-feel-like-the-bot-ignores-rapidfire-messages)
- [Common troubleshooting](#common-troubleshooting)
-  - [“All models failed” — what should I check first?](#all-models-failed-what-should-i-check-first)
-  - [I’m running on my personal WhatsApp number — why is self-chat weird?](#im-running-on-my-personal-whatsapp-number-why-is-self-chat-weird)
-  - [WhatsApp logged me out. How do I re‑auth?](#whatsapp-logged-me-out-how-do-i-reauth)
-  - [Build errors on `main` — what’s the standard fix path?](#build-errors-on-main-whats-the-standard-fix-path)
-  - [npm install fails (allow-build-scripts / missing tar or yargs). What now?](#npm-install-fails-allow-build-scripts--missing-tar-or-yargs-what-now)
-  - [How do I switch between git installs and npm installs?](#how-do-i-switch-between-git-installs-and-npm-installs)
-  - [Telegram block streaming isn’t splitting text between tool calls. Why?](#telegram-block-streaming-isnt-splitting-text-between-tool-calls-why)
-  - [Discord doesn’t reply in my server even with `requireMention: false`. Why?](#discord-doesnt-reply-in-my-server-even-with-requiremention-false-why)
-  - [Cloud Code Assist API error: invalid tool schema (400). What now?](#cloud-code-assist-api-error-invalid-tool-schema-400-what-now)

 ## First 60 seconds if something's broken

@@ -858,7 +853,21 @@ See [/concepts/model-providers](/concepts/model-providers) and [/environment](/e

 Send `/new` or `/reset` as a standalone message. See [Session management](/concepts/session).

-### How do I completely reset Clawdbot (but keep it installed)?
+### Do sessions reset automatically if I never send `/new`?
+
+Yes. Sessions expire after `session.idleMinutes` (default **60**). The **next**
+message starts a fresh session id for that chat key. This does not delete
+transcripts — it just starts a new session.
+
+```json5
+{
+  session: {
+    idleMinutes: 240
+  }
+}
+```
+
+### How do I completely reset Clawdbot but keep it installed?

 Use the reset command:

@@ -905,6 +914,24 @@ If it keeps happening:

 Docs: [Compaction](/concepts/compaction), [Session pruning](/concepts/session-pruning), [Session management](/concepts/session).

+### Why am I getting heartbeat messages every 30 minutes?
+
+Heartbeats run every **30m** by default. Tune or disable them:
+
+```json5
+{
+  agents: {
+    defaults: {
+      heartbeat: {
+        every: "2h"   // or "0m" to disable
+      }
+    }
+  }
+}
+```
+
+Per-agent overrides use `agents.list[].heartbeat`. Docs: [Heartbeat](/gateway/heartbeat).
+
 ### Do I need to add a “bot account” to a WhatsApp group?

 No. Clawdbot runs on **your own account**, so if you’re in the group, Clawdbot can see it.
@@ -935,6 +962,19 @@ See [Groups](/concepts/groups) and [Group messages](/concepts/group-messages).

 Direct chats collapse to the main session by default. Groups/channels have their own session keys, and Telegram topics / Discord threads are separate sessions. See [Groups](/concepts/groups) and [Group messages](/concepts/group-messages).

+### How many workspaces and agents can I create?
+
+No hard limits. Dozens (even hundreds) are fine, but watch for:
+
+- **Disk growth:** sessions + transcripts live under `~/.clawdbot/agents/<agentId>/sessions/`.
+- **Token cost:** more agents means more concurrent model usage.
+- **Ops overhead:** per-agent auth profiles, workspaces, and channel routing.
+
+Tips:
+- Keep one **active** workspace per agent (`agents.defaults.workspace`).
+- Prune old sessions (delete JSONL or store entries) if disk grows.
+- Use `clawdbot doctor` to spot stray workspaces and profile mismatches.
+
 ## Models: defaults, selection, aliases, switching

 ### What is the “default model”?
@@ -1315,9 +1355,11 @@ Facts (from code):

 Fix:
 - Fastest: `clawdbot dashboard` (prints + copies tokenized link, tries to open; shows SSH hint if headless).
+- If you don’t have a token yet: `clawdbot doctor --generate-gateway-token`.
 - If remote, tunnel first: `ssh -N -L 18789:127.0.0.1:18789 user@host` then open `http://127.0.0.1:18789/?token=...`.
 - Set `gateway.auth.token` (or `CLAWDBOT_GATEWAY_TOKEN`) on the gateway host.
 - In the Control UI settings, paste the same token (or refresh with a one-time `?token=...` link).
+- Still stuck? Run `clawdbot status --all` and follow [Troubleshooting](/gateway/troubleshooting). See [Dashboard](/web/dashboard) for auth details.

 ### I set `gateway.bind: "tailnet"` but it can’t bind / nothing listens

@@ -1407,6 +1449,14 @@ clawdbot daemon restart

 If you run the gateway manually, `clawdbot gateway --force` can reclaim the port. See [Gateway](/gateway).

+### ELI5: `clawdbot daemon restart` vs `clawdbot gateway`
+
+- `clawdbot daemon restart`: restarts the **background service** (launchd/systemd).
+- `clawdbot gateway`: runs the gateway **in the foreground** for this terminal session.
+
+If you installed the daemon, use the daemon commands. Use `clawdbot gateway` when
+you want a one-off, foreground run.
+
 ### What’s the fastest way to get more details when something fails?

 Start the Gateway with `--verbose` to get more console detail. Then inspect the log file for channel auth, model routing, and RPC errors.
@@ -1423,7 +1473,11 @@ CLI sending:
 clawdbot message send --to +15555550123 --message "Here you go" --media /path/to/file.png
 ```

-Note: images are resized/recompressed (max side 2048px) to hit size limits. See [Images](/nodes/images).
+Also check:
+- The target channel supports outbound media and isn’t blocked by allowlists.
+- The file is within the provider’s size limits (images are resized to max 2048px).
+
+See [Images](/nodes/images).

 ## Security and access control

@@ -1439,6 +1493,19 @@ Treat inbound DMs as untrusted input. Defaults are designed to reduce risk:

 Run `clawdbot doctor` to surface risky DM policies.

+### I ran `/start` in Telegram but didn’t get a pairing code
+
+Pairing codes are sent **only** when an unknown sender messages the bot and
+`dmPolicy: "pairing"` is enabled. `/start` by itself doesn’t generate a code.
+
+Check pending requests:
+```bash
+clawdbot pairing list telegram
+```
+
+If you want immediate access, allowlist your sender id or set `dmPolicy: "open"`
+for that account.
+
 ### WhatsApp: will it message my contacts? How does pairing work?

 No. Default WhatsApp DM policy is **pairing**. Unknown senders only get a pairing code and their message is **not processed**. Clawdbot only replies to chats it receives or to explicit sends you trigger.
@@ -1496,139 +1563,6 @@ Queue mode controls how new messages interact with an in‑flight run. Use `/que

 You can add options like `debounce:2s cap:25 drop:summarize` for followup modes.

-## Common troubleshooting
-
-### “All models failed” — what should I check first?
-
- **Credentials** present for the provider(s) being tried (auth profiles + env vars).
- **Model routing**: confirm `agents.defaults.model.primary` and fallbacks are models you can access.
- **Gateway logs** in `/tmp/clawdbot/…` for the exact provider error.
- **`/model status`** to see current configured models + shorthands.
-
-### I’m running on my personal WhatsApp number — why is self-chat weird?
-
-Enable self-chat mode and allowlist your own number:
-
-```json5
-{
-  channels: {
-    whatsapp: {
-      selfChatMode: true,
-      dmPolicy: "allowlist",
-      allowFrom: ["+15555550123"]
-    }
-  }
-}
-```
-
-See [WhatsApp setup](/channels/whatsapp).
-
-### WhatsApp logged me out. How do I re‑auth?
-
-Run the login command again and scan the QR code:
-
-```bash
-clawdbot channels login
-```
-
-### Build errors on `main` — what’s the standard fix path?
-
-1) `git pull origin main && pnpm install`
-2) `pnpm clawdbot doctor`
-3) Check GitHub issues or Discord
-4) Temporary workaround: check out an older commit
-
-### npm install fails (allow-build-scripts / missing tar or yargs). What now?
-
-If you’re running from source, use the repo’s package manager: **pnpm** (preferred).
-The repo declares `packageManager: "pnpm@…"`, and pnpm patches are tracked in `pnpm.patchedDependencies`.
-
-Typical recovery:
-```bash
-git status   # ensure you’re in the repo root
-pnpm install
-pnpm build
-pnpm clawdbot doctor
-clawdbot daemon restart
-```
-
-Why: pnpm is the configured package manager for this repo, and the dependency
-patching workflow relies on it.
-
-### How do I switch between git installs and npm installs?
-
-Use the **website installer** and select the install method with a flag. It
-upgrades in place and rewrites the gateway service to point at the new install.
-
-Switch **to git install**:
-```bash
-curl -fsSL https://clawd.bot/install.sh | bash -s -- --install-method git --no-onboard
-```
-
-Switch **to npm global**:
-```bash
-curl -fsSL https://clawd.bot/install.sh | bash
-```
-
-Notes:
- The git flow only rebases if the repo is clean. Commit or stash changes first.
- After switching, run:
-  ```bash
-  clawdbot doctor
-  clawdbot daemon restart
-  ```
-
-### Telegram block streaming isn’t splitting text between tool calls. Why?
-
-Block streaming only sends **completed text blocks**. Common reasons you see a single message:
- `agents.defaults.blockStreamingDefault` is still `"off"`.
- `channels.telegram.blockStreaming` is set to `false`.
- `channels.telegram.streamMode` is `partial` or `block` **and draft streaming is active**
-  (private chat + topics). Draft streaming disables block streaming in that case.
- Your `minChars` / coalesce settings are too high, so chunks get merged.
- The model emits one large text block (no mid‑reply flush points).
-
-Fix checklist:
-1) Put block streaming settings under `agents.defaults`, not the root.
-2) Set `channels.telegram.streamMode: "off"` if you want real multi‑message block replies.
-3) Use smaller chunk/coalesce thresholds while debugging.
-
-See [Streaming](/concepts/streaming).
-
-### Discord doesn’t reply in my server even with `requireMention: false`. Why?
-
-`requireMention` only controls mention‑gating **after** the channel passes allowlists.
-By default `channels.discord.groupPolicy` is **allowlist**, so guilds must be explicitly enabled.
-If you set `channels.discord.guilds.<guildId>.channels`, only the listed channels are allowed; omit it to allow all channels in the guild.
-
-Fix checklist:
-1) Set `channels.discord.groupPolicy: "open"` **or** add a guild allowlist entry (and optionally a channel allowlist).
-2) Use **numeric channel IDs** in `channels.discord.guilds.<guildId>.channels`.
-3) Put `requireMention: false` **under** `channels.discord.guilds` (global or per‑channel).
-   Top‑level `channels.discord.requireMention` is not a supported key.
-4) Ensure the bot has **Message Content Intent** and channel permissions.
-5) Run `clawdbot channels status --probe` for audit hints.
-
-Docs: [Discord](/channels/discord), [Channels troubleshooting](/channels/troubleshooting).
-
-### Cloud Code Assist API error: invalid tool schema (400). What now?
-
-This is almost always a **tool schema compatibility** issue. The Cloud Code Assist
-endpoint accepts a strict subset of JSON Schema. Clawdbot scrubs/normalizes tool
-schemas in current `main`, but the fix is not in the last release yet (as of
-January 13, 2026).
-
-Fix checklist:
-1) **Update Clawdbot**:
-   - If you can run from source, pull `main` and restart the gateway.
-   - Otherwise, wait for the next release that includes the schema scrubber.
-2) Avoid unsupported keywords like `anyOf/oneOf/allOf`, `patternProperties`,
-   `additionalProperties`, `minLength`, `maxLength`, `format`, etc.
-3) If you define custom tools, keep the top‑level schema as `type: "object"` with
-   `properties` and simple enums.
-
-See [Tools](/tools) and [TypeBox schemas](/concepts/typebox).
-
 ## Answer the exact question from the screenshot/chat log

 **Q: “What’s the default model for Anthropic with an API key?”**
@@ -68,6 +68,8 @@ Tip: `--json` does **not** imply non-interactive mode. Use `--non-interactive` (

 1) **Existing config detection**
   - If `~/.clawdbot/clawdbot.json` exists, choose **Keep / Modify / Reset**.
+   - Re-running the wizard does **not** wipe anything unless you explicitly choose **Reset**
+     (or pass `--reset`).
   - If the config is invalid or contains legacy keys, the wizard stops and asks
     you to run `clawdbot doctor` before continuing.
   - Reset uses `trash` (never `rm`) and offers scopes:
@@ -112,3 +112,18 @@ curl -s http://127.0.0.1:18791/tabs
 | `browser.noSandbox` | Add `--no-sandbox` flag (needed for some Linux setups) | `false` |
 | `browser.attachOnly` | Don't launch browser, only attach to existing | `false` |
 | `browser.cdpPort` | Chrome DevTools Protocol port | `18800` |
+
+### Problem: "Chrome extension relay is running, but no tab is connected"
+
+You’re using the `chrome` profile (extension relay). It expects the Clawdbot
+browser extension to be attached to a live tab.
+
+Fix options:
+1. **Use the managed browser:** `clawdbot browser start --browser-profile clawd`
+   (or set `browser.defaultProfile: "clawd"`).
+2. **Use the extension relay:** install the extension, open a tab, and click the
+   Clawdbot extension icon to attach it.
+
+Notes:
+- The `chrome` profile uses your **system default Chromium browser** when possible.
+- Local `clawd` profiles auto-assign `cdpPort`/`cdpUrl`; only set those for remote CDP.
@@ -14,8 +14,10 @@ control server.

 Beginner view:
 - Think of it as a **separate, agent-only browser**.
- It does **not** touch your personal browser profile.
+- The `clawd` profile does **not** touch your personal browser profile.
 - The agent can **open tabs, read pages, click, and type** in a safe lane.
+- The default `chrome` profile uses the **system default Chromium browser** via the
+  extension relay; switch to `clawd` for the isolated managed browser.

 ## What you get

@@ -39,6 +41,14 @@ clawdbot browser --browser-profile clawd snapshot
 If you get “Browser disabled”, enable it in config (see below) and restart the
 Gateway.

+## Profiles: `clawd` vs `chrome`
+
+- `clawd`: managed, isolated browser (no extension required).
+- `chrome`: extension relay to your **system browser** (requires the Clawdbot
+  extension to be attached to a tab).
+
+Set `browser.defaultProfile: "clawd"` if you want managed mode by default.
+
 ## Configuration

 Browser settings live in `~/.clawdbot/clawdbot.json`.
@@ -49,6 +59,8 @@ Browser settings live in `~/.clawdbot/clawdbot.json`.
    enabled: true,                    // default: true
    controlUrl: "http://127.0.0.1:18791",
    cdpUrl: "http://127.0.0.1:18792", // defaults to controlUrl + 1
+    remoteCdpTimeoutMs: 1500,         // remote CDP HTTP timeout (ms)
+    remoteCdpHandshakeTimeoutMs: 3000, // remote CDP WebSocket handshake timeout (ms)
    defaultProfile: "chrome",
    color: "#FF4500",
    headless: false,
@@ -69,13 +81,25 @@ Notes:
 - If you override the Gateway port (`gateway.port` or `CLAWDBOT_GATEWAY_PORT`),
  the default browser ports shift to stay in the same “family” (control = gateway + 2).
 - `cdpUrl` defaults to `controlUrl + 1` when unset.
+- `remoteCdpTimeoutMs` applies to remote (non-loopback) CDP reachability checks.
+- `remoteCdpHandshakeTimeoutMs` applies to remote CDP WebSocket reachability checks.
 - `attachOnly: true` means “never launch a local browser; only attach if it is already running.”
 - `color` + per-profile `color` tint the browser UI so you can see which profile is active.
- Auto-detect order: default browser if Chromium-based; otherwise Chrome → Brave → Edge → Chromium → Chrome Canary.
+- Default profile is `chrome` (extension relay). Use `defaultProfile: "clawd"` for the managed browser.
+- Auto-detect order: system default browser if Chromium-based; otherwise Chrome → Brave → Edge → Chromium → Chrome Canary.
+- Local `clawd` profiles auto-assign `cdpPort`/`cdpUrl` — set those only for remote CDP.

 ## Use Brave (or another Chromium-based browser)

-Set `browser.executablePath` to override auto-detection:
+If your **system default** browser is Chromium-based (Chrome/Brave/Edge/etc),
+Clawdbot uses it automatically. Set `browser.executablePath` to override
+auto-detection:
+
+CLI example:
+
+```bash
+clawdbot config set browser.executablePath "/usr/bin/google-chrome"
+```

 ```json5
 // macOS
@@ -134,6 +158,42 @@ Example:
 Use `profiles.<name>.cdpUrl` for **remote CDP** if you want the Gateway to talk
 directly to a Chromium-based browser instance without a remote control server.

+Remote CDP URLs can include auth:
+- Query tokens (e.g., `https://provider.example?token=<token>`)
+- HTTP Basic auth (e.g., `https://user:pass@provider.example`)
+
+Clawdbot preserves the auth when calling `/json/*` endpoints and when connecting
+to the CDP WebSocket. Prefer environment variables or secrets managers for
+tokens instead of committing them to config files.
+
+### Browserless (hosted remote CDP)
+
+[Browserless](https://browserless.io) is a hosted Chromium service that exposes
+CDP endpoints over HTTPS. You can point a Clawdbot browser profile at a
+Browserless region endpoint and authenticate with your API key.
+
+Example:
+```json5
+{
+  browser: {
+    enabled: true,
+    defaultProfile: "browserless",
+    remoteCdpTimeoutMs: 2000,
+    remoteCdpHandshakeTimeoutMs: 4000,
+    profiles: {
+      browserless: {
+        cdpUrl: "https://production-sfo.browserless.io?token=<BROWSERLESS_API_KEY>",
+        color: "#00AA00"
+      }
+    }
+  }
+}
+```
+
+Notes:
+- Replace `<BROWSERLESS_API_KEY>` with your real Browserless token.
+- Choose the region endpoint that matches your Browserless account (see their docs).
+
 ### Running the control server on the browser machine

 Run a standalone browser control server (recommended when your Gateway is remote):
@@ -339,6 +399,10 @@ Playwright. If Playwright isn’t installed, those endpoints return a clear 501
 error. ARIA snapshots and basic screenshots still work for clawd-managed Chrome.
 For the Chrome extension relay driver, ARIA snapshots and screenshots require Playwright.

+If you see `Playwright is not available in this gateway build`, install the full
+Playwright package (not `playwright-core`) and restart the gateway, or reinstall
+Clawdbot with browser support.
+
 ## How it works (internal)

 High-level flow:
@@ -206,6 +206,31 @@ if [[ "$NO_SIGN" -ne 1 && -f "${LAUNCHAGENT_DISABLE_MARKER}" ]]; then
  run_step "clear launchagent disable marker" /bin/rm -f "${LAUNCHAGENT_DISABLE_MARKER}"
 fi

+# When unsigned, ensure the gateway LaunchAgent targets the repo CLI (before the app launches).
+# This reduces noisy "could not connect" errors during app startup.
+if [ "$NO_SIGN" -eq 1 ]; then
+  run_step "install gateway launch agent (unsigned)" bash -lc "cd '${ROOT_DIR}' && node dist/entry.js daemon install --force --runtime node"
+  run_step "restart gateway daemon (unsigned)" bash -lc "cd '${ROOT_DIR}' && node dist/entry.js daemon restart"
+  if [[ "${GATEWAY_WAIT_SECONDS}" -gt 0 ]]; then
+    run_step "wait for gateway (unsigned)" sleep "${GATEWAY_WAIT_SECONDS}"
+  fi
+  GATEWAY_PORT="$(
+    node -e '
+      const fs = require("node:fs");
+      const path = require("node:path");
+      try {
+        const raw = fs.readFileSync(path.join(process.env.HOME, ".clawdbot", "clawdbot.json"), "utf8");
+        const cfg = JSON.parse(raw);
+        const port = cfg && cfg.gateway && typeof cfg.gateway.port === "number" ? cfg.gateway.port : 18789;
+        process.stdout.write(String(port));
+      } catch {
+        process.stdout.write("18789");
+      }
+    '
+  )"
+  run_step "verify gateway port ${GATEWAY_PORT} (unsigned)" bash -lc "lsof -iTCP:${GATEWAY_PORT} -sTCP:LISTEN | head -n 5 || true"
+fi
+
 # 4) Launch the installed app in the foreground so the menu bar extra appears.
 # LaunchServices can inherit a huge environment from this shell (secrets, prompt vars, etc.).
 # That can cause launchd spawn failures and is undesirable for a GUI app anyway.
@@ -226,13 +251,6 @@ else
  fail "App exited immediately. Check ${LOG_PATH} or Console.app (User Reports)."
 fi

-# When unsigned, ensure the gateway LaunchAgent targets the repo CLI (after the app launches).
 if [ "$NO_SIGN" -eq 1 ]; then
-  run_step "install gateway launch agent (unsigned)" bash -lc "cd '${ROOT_DIR}' && node dist/entry.js daemon install --force --runtime node"
-  run_step "restart gateway daemon (unsigned)" bash -lc "cd '${ROOT_DIR}' && node dist/entry.js daemon restart"
-  if [[ "${GATEWAY_WAIT_SECONDS}" -gt 0 ]]; then
-    run_step "wait for gateway (unsigned)" sleep "${GATEWAY_WAIT_SECONDS}"
-  fi
-  run_step "verify gateway port 18789 (unsigned)" bash -lc "lsof -iTCP:18789 -sTCP:LISTEN | head -n 5 || true"
  run_step "show gateway launch agent args (unsigned)" bash -lc "/usr/bin/plutil -p '${HOME}/Library/LaunchAgents/com.clawdbot.gateway.plist' | head -n 40 || true"
 fi
@@ -19,13 +19,53 @@ open ~/Projects/tmp/openai-image-gen-*/index.html  # if ~/Projects/tmp exists; e
 Useful flags:

 ```bash
+# GPT image models with various options
 python3 {baseDir}/scripts/gen.py --count 16 --model gpt-image-1
 python3 {baseDir}/scripts/gen.py --prompt "ultra-detailed studio photo of a lobster astronaut" --count 4
 python3 {baseDir}/scripts/gen.py --size 1536x1024 --quality high --out-dir ./out/images
+python3 {baseDir}/scripts/gen.py --model gpt-image-1.5 --background transparent --output-format webp
+
+# DALL-E 3 (note: count is automatically limited to 1)
+python3 {baseDir}/scripts/gen.py --model dall-e-3 --quality hd --size 1792x1024 --style vivid
+python3 {baseDir}/scripts/gen.py --model dall-e-3 --style natural --prompt "serene mountain landscape"
+
+# DALL-E 2
+python3 {baseDir}/scripts/gen.py --model dall-e-2 --size 512x512 --count 4
 ```

+## Model-Specific Parameters
+
+Different models support different parameter values. The script automatically selects appropriate defaults based on the model.
+
+### Size
+
+- **GPT image models** (`gpt-image-1`, `gpt-image-1-mini`, `gpt-image-1.5`): `1024x1024`, `1536x1024` (landscape), `1024x1536` (portrait), or `auto`
+  - Default: `1024x1024`
+- **dall-e-3**: `1024x1024`, `1792x1024`, or `1024x1792`
+  - Default: `1024x1024`
+- **dall-e-2**: `256x256`, `512x512`, or `1024x1024`
+  - Default: `1024x1024`
+
+### Quality
+
+- **GPT image models**: `auto`, `high`, `medium`, or `low`
+  - Default: `high`
+- **dall-e-3**: `hd` or `standard`
+  - Default: `standard`
+- **dall-e-2**: `standard` only
+  - Default: `standard`
+
+### Other Notable Differences
+
+- **dall-e-3** only supports generating 1 image at a time (`n=1`). The script automatically limits count to 1 when using this model.
+- **GPT image models** support additional parameters:
+  - `--background`: `transparent`, `opaque`, or `auto` (default)
+  - `--output-format`: `png` (default), `jpeg`, or `webp`
+  - Note: `stream` and `moderation` are available via API but not yet implemented in this script
+- **dall-e-3** has a `--style` parameter: `vivid` (hyper-real, dramatic) or `natural` (more natural looking)
+
 ## Output

- `*.png` images
+- `*.png`, `*.jpeg`, or `*.webp` images (output format depends on model + `--output-format`)
 - `prompts.json` (prompt → file mapping)
 - `index.html` (thumbnail gallery)
@@ -62,24 +62,53 @@ def pick_prompts(count: int) -> list[str]:
    return prompts


+def get_model_defaults(model: str) -> tuple[str, str]:
+    """Return (default_size, default_quality) for the given model."""
+    if model == "dall-e-2":
+        # quality will be ignored
+        return ("1024x1024", "standard")
+    elif model == "dall-e-3":
+        return ("1024x1024", "standard")
+    else:
+        # GPT image or future models
+        return ("1024x1024", "high")
+
+
 def request_images(
    api_key: str,
    prompt: str,
    model: str,
    size: str,
    quality: str,
+    background: str = "",
+    output_format: str = "",
+    style: str = "",
 ) -> dict:
    url = "https://api.openai.com/v1/images/generations"
-    body = json.dumps(
-        {
-            "model": model,
-            "prompt": prompt,
-            "size": size,
-            "quality": quality,
-            "n": 1,
-            "response_format": "b64_json",
-        }
-    ).encode("utf-8")
+    args = {
+        "model": model,
+        "prompt": prompt,
+        "size": size,
+        "n": 1,
+    }
+
+    # Quality parameter - dall-e-2 doesn't accept this parameter
+    if model != "dall-e-2":
+        args["quality"] = quality
+
+    if model.startswith("dall-e"):
+        args["response_format"] = "b64_json"
+
+    if model.startswith("gpt-image"):
+        if background:
+            args["background"] = background
+        if output_format:
+            args["output_format"] = output_format
+
+    if model == "dall-e-3" and style:
+        args["style"] = style
+
+    body = json.dumps(args).encode("utf-8")
    req = urllib.request.Request(
        url,
        method="POST",
@@ -136,8 +165,11 @@ def main() -> int:
    ap.add_argument("--prompt", help="Single prompt. If omitted, random prompts are generated.")
    ap.add_argument("--count", type=int, default=8, help="How many images to generate.")
    ap.add_argument("--model", default="gpt-image-1", help="Image model id.")
-    ap.add_argument("--size", default="1024x1024", help="Image size (e.g. 1024x1024, 1536x1024).")
-    ap.add_argument("--quality", default="high", help="Image quality (varies by model).")
+    ap.add_argument("--size", default="", help="Image size (e.g. 1024x1024, 1536x1024). Defaults based on model if not specified.")
+    ap.add_argument("--quality", default="", help="Image quality (e.g. high, standard). Defaults based on model if not specified.")
+    ap.add_argument("--background", default="", help="Background transparency (GPT models only): transparent, opaque, or auto.")
+    ap.add_argument("--output-format", default="", help="Output format (GPT models only): png, jpeg, or webp.")
+    ap.add_argument("--style", default="", help="Image style (dall-e-3 only): vivid or natural.")
    ap.add_argument("--out-dir", default="", help="Output directory (default: ./tmp/openai-image-gen-<ts>).")
    args = ap.parse_args()

@@ -146,21 +178,46 @@ def main() -> int:
        print("Missing OPENAI_API_KEY", file=sys.stderr)
        return 2

+    # Apply model-specific defaults if not specified
+    default_size, default_quality = get_model_defaults(args.model)
+    size = args.size or default_size
+    quality = args.quality or default_quality
+
+    count = args.count
+    if args.model == "dall-e-3" and count > 1:
+        print(f"Warning: dall-e-3 only supports generating 1 image at a time. Reducing count from {count} to 1.", file=sys.stderr)
+        count = 1
+
    out_dir = Path(args.out_dir).expanduser() if args.out_dir else default_out_dir()
    out_dir.mkdir(parents=True, exist_ok=True)

-    prompts = [args.prompt] * args.count if args.prompt else pick_prompts(args.count)
+    prompts = [args.prompt] * count if args.prompt else pick_prompts(count)
+
+    # Determine file extension based on output format
+    if args.model.startswith("gpt-image") and args.output_format:
+        file_ext = args.output_format
+    else:
+        file_ext = "png"

    items: list[dict] = []
    for idx, prompt in enumerate(prompts, start=1):
        print(f"[{idx}/{len(prompts)}] {prompt}")
-        res = request_images(api_key, prompt, args.model, args.size, args.quality)
+        res = request_images(
+            api_key,
+            prompt,
+            args.model,
+            size,
+            quality,
+            args.background,
+            args.output_format,
+            args.style,
+        )
        b64 = res.get("data", [{}])[0].get("b64_json")
        if not b64:
            raise RuntimeError(f"Unexpected response: {json.dumps(res)[:400]}")
-        png = base64.b64decode(b64)
-        filename = f"{idx:03d}-{slugify(prompt)[:40]}.png"
-        (out_dir / filename).write_bytes(png)
+        image_bytes = base64.b64decode(b64)
+        filename = f"{idx:03d}-{slugify(prompt)[:40]}.{file_ext}"
+        (out_dir / filename).write_bytes(image_bytes)
        items.append({"prompt": prompt, "file": filename})

    (out_dir / "prompts.json").write_text(json.dumps(items, indent=2), encoding="utf-8")
@@ -99,6 +99,7 @@ describe("session_status tool", () => {
    expect(details.ok).toBe(true);
    expect(details.statusText).toContain("Clawdbot");
    expect(details.statusText).toContain("🧠 Model:");
+    expect(details.statusText).not.toContain("OAuth/token status");
  });

  it("errors for unknown session keys", async () => {
@@ -75,7 +75,9 @@ export function createClawdbotTools(options?: {
    }),
    createCanvasTool(),
    createNodesTool(),
-    createCronTool(),
+    createCronTool({
+      agentSessionKey: options?.agentSessionKey,
+    }),
    createMessageTool({
      agentAccountId: options?.agentAccountId,
      config: options?.config,
@@ -9,26 +9,63 @@ import type { EmbeddedContextFile } from "./types.js";

 type ContentBlockWithSignature = {
  thought_signature?: unknown;
+  thoughtSignature?: unknown;
  [key: string]: unknown;
 };

+type ThoughtSignatureSanitizeOptions = {
+  allowBase64Only?: boolean;
+  includeCamelCase?: boolean;
+};
+
+function isBase64Signature(value: string): boolean {
+  const trimmed = value.trim();
+  if (!trimmed) return false;
+  const compact = trimmed.replace(/\s+/g, "");
+  if (!/^[A-Za-z0-9+/=_-]+$/.test(compact)) return false;
+  const isUrl = compact.includes("-") || compact.includes("_");
+  try {
+    const buf = Buffer.from(compact, isUrl ? "base64url" : "base64");
+    if (buf.length === 0) return false;
+    const encoded = buf.toString(isUrl ? "base64url" : "base64");
+    const normalize = (input: string) => input.replace(/=+$/g, "");
+    return normalize(encoded) === normalize(compact);
+  } catch {
+    return false;
+  }
+}
+
 /**
 * Strips Claude-style thought_signature fields from content blocks.
 *
 * Gemini expects thought signatures as base64-encoded bytes, but Claude stores message ids
 * like "msg_abc123...". We only strip "msg_*" to preserve any provider-valid signatures.
 */
-export function stripThoughtSignatures<T>(content: T): T {
+export function stripThoughtSignatures<T>(
+  content: T,
+  options?: ThoughtSignatureSanitizeOptions,
+): T {
  if (!Array.isArray(content)) return content;
+  const allowBase64Only = options?.allowBase64Only ?? false;
+  const includeCamelCase = options?.includeCamelCase ?? false;
+  const shouldStripSignature = (value: unknown): boolean => {
+    if (!allowBase64Only) {
+      return typeof value === "string" && value.startsWith("msg_");
+    }
+    return typeof value !== "string" || !isBase64Signature(value);
+  };
  return content.map((block) => {
    if (!block || typeof block !== "object") return block;
    const rec = block as ContentBlockWithSignature;
-    const signature = rec.thought_signature;
-    if (typeof signature !== "string" || !signature.startsWith("msg_")) {
+    const stripSnake = shouldStripSignature(rec.thought_signature);
+    const stripCamel = includeCamelCase ? shouldStripSignature(rec.thoughtSignature) : false;
+    if (!stripSnake && !stripCamel) {
      return block;
    }
-    const { thought_signature: _signature, ...rest } = rec;
-    return rest;
+    const next = { ...rec };
+    if (stripSnake) delete next.thought_signature;
+    if (stripCamel) delete next.thoughtSignature;
+    return next;
  }) as T;
 }

@@ -23,6 +23,7 @@ export { sanitizeGoogleTurnOrdering };
 type GeminiToolCallBlock = {
  type?: unknown;
  thought_signature?: unknown;
+  thoughtSignature?: unknown;
  id?: unknown;
  toolCallId?: unknown;
  name?: unknown;
@@ -118,7 +119,8 @@ export function downgradeGeminiHistory(messages: AgentMessage[]): AgentMessage[]
        const blockRecord = block as GeminiToolCallBlock;
        const type = blockRecord.type;
        if (type === "toolCall" || type === "functionCall" || type === "toolUse") {
-          const hasSignature = Boolean(blockRecord.thought_signature);
+          const signature = blockRecord.thought_signature ?? blockRecord.thoughtSignature;
+          const hasSignature = Boolean(signature);
          if (!hasSignature) {
            const id =
              typeof blockRecord.id === "string"
@@ -34,6 +34,10 @@ export async function sanitizeSessionMessagesImages(
    sanitizeToolCallIds?: boolean;
    enforceToolCallLast?: boolean;
    preserveSignatures?: boolean;
+    sanitizeThoughtSignatures?: {
+      allowBase64Only?: boolean;
+      includeCamelCase?: boolean;
+    };
  },
 ): Promise<AgentMessage[]> {
  // We sanitize historical session messages because Anthropic can reject a request
@@ -82,7 +86,7 @@ export async function sanitizeSessionMessagesImages(
      if (Array.isArray(content)) {
        const strippedContent = options?.preserveSignatures
          ? content // Keep signatures for Antigravity Claude
-          : stripThoughtSignatures(content); // Strip for Gemini
+          : stripThoughtSignatures(content, options?.sanitizeThoughtSignatures); // Strip for Gemini

        const filteredContent = strippedContent.filter((block) => {
          if (!block || typeof block !== "object") return true;
@@ -145,6 +145,65 @@ describe("sanitizeSessionHistory (google thinking)", () => {
    expect(assistant.content?.[1]?.text).toBe("internal note");
  });

+  it("strips non-base64 thought signatures for OpenRouter Gemini", async () => {
+    const sessionManager = SessionManager.inMemory();
+    const input = [
+      {
+        role: "user",
+        content: "hi",
+      },
+      {
+        role: "assistant",
+        content: [
+          { type: "text", text: "hello", thought_signature: "msg_abc123" },
+          { type: "thinking", thinking: "ok", thought_signature: "c2ln" },
+          {
+            type: "toolCall",
+            id: "call_1",
+            name: "read",
+            arguments: { path: "/tmp/foo" },
+            thoughtSignature: '{"id":1}',
+          },
+          {
+            type: "toolCall",
+            id: "call_2",
+            name: "read",
+            arguments: { path: "/tmp/bar" },
+            thoughtSignature: "c2ln",
+          },
+        ],
+      },
+    ] satisfies AgentMessage[];
+
+    const out = await sanitizeSessionHistory({
+      messages: input,
+      modelApi: "openrouter",
+      provider: "openrouter",
+      modelId: "google/gemini-1.5-pro",
+      sessionManager,
+      sessionId: "session:openrouter-gemini",
+    });
+
+    const assistant = out.find((msg) => (msg as { role?: string }).role === "assistant") as {
+      content?: Array<{ type?: string; thought_signature?: string; thoughtSignature?: string }>;
+    };
+    expect(assistant.content).toEqual([
+      { type: "text", text: "hello" },
+      { type: "text", text: "ok" },
+      {
+        type: "text",
+        text: '[Tool Call: read (ID: call_1)]\nArguments: {\n  "path": "/tmp/foo"\n}',
+      },
+      {
+        type: "toolCall",
+        id: "call_2",
+        name: "read",
+        arguments: { path: "/tmp/bar" },
+        thoughtSignature: "c2ln",
+      },
+    ]);
+  });
+
  it("downgrades only unsigned thinking blocks when mixed with signed ones", async () => {
    const sessionManager = SessionManager.inMemory();
    const input = [
@@ -261,4 +261,64 @@ describe("runEmbeddedPiAgent", () => {
    expect(secondUserIndex).toBeGreaterThan(firstAssistantIndex);
    expect(secondAssistantIndex).toBeGreaterThan(secondUserIndex);
  }, 20_000);
+  it("returns role ordering error when session ends with a user turn", async () => {
+    const { SessionManager } = await import("@mariozechner/pi-coding-agent");
+
+    const agentDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-agent-"));
+    const workspaceDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-workspace-"));
+    const sessionFile = path.join(workspaceDir, "session.jsonl");
+
+    const sessionManager = SessionManager.open(sessionFile);
+    sessionManager.appendMessage({
+      role: "user",
+      content: [{ type: "text", text: "seed user 1" }],
+    });
+    sessionManager.appendMessage({
+      role: "assistant",
+      content: [{ type: "text", text: "seed assistant" }],
+      stopReason: "stop",
+      api: "openai-responses",
+      provider: "openai",
+      model: "mock-1",
+      usage: {
+        input: 1,
+        output: 1,
+        cacheRead: 0,
+        cacheWrite: 0,
+        totalTokens: 2,
+        cost: {
+          input: 0,
+          output: 0,
+          cacheRead: 0,
+          cacheWrite: 0,
+          total: 0,
+        },
+      },
+      timestamp: Date.now(),
+    });
+    sessionManager.appendMessage({
+      role: "user",
+      content: [{ type: "text", text: "seed user 2" }],
+    });
+
+    const cfg = makeOpenAiConfig(["mock-1"]);
+    await ensureModels(cfg, agentDir);
+
+    const result = await runEmbeddedPiAgent({
+      sessionId: "session:test",
+      sessionKey: "agent:main:main",
+      sessionFile,
+      workspaceDir,
+      config: cfg,
+      prompt: "hello",
+      provider: "openai",
+      model: "mock-1",
+      timeoutMs: 5_000,
+      agentDir,
+    });
+
+    expect(result.meta.error?.kind).toBe("role_ordering");
+    expect(result.meta.error?.message).toMatch(/incorrect role information|roles must alternate/i);
+    expect(result.payloads?.[0]?.text).toContain("Message ordering conflict");
+  });
 });
@@ -0,0 +1,105 @@
+import type { AgentMessage } from "@mariozechner/pi-agent-core";
+import type { SessionManager } from "@mariozechner/pi-coding-agent";
+import { beforeEach, describe, expect, it, vi } from "vitest";
+import * as helpers from "./pi-embedded-helpers.js";
+import { sanitizeSessionHistory } from "./pi-embedded-runner/google.js";
+
+// Mock dependencies
+vi.mock("./pi-embedded-helpers.js", async () => {
+  const actual = await vi.importActual("./pi-embedded-helpers.js");
+  return {
+    ...actual,
+    isGoogleModelApi: vi.fn(),
+    downgradeGeminiHistory: vi.fn(),
+    sanitizeSessionMessagesImages: vi.fn().mockImplementation(async (msgs) => msgs),
+  };
+});
+
+// We don't mock session-transcript-repair.js as it is a pure function and complicates mocking.
+// We rely on the real implementation which should pass through our simple messages.
+
+describe("sanitizeSessionHistory", () => {
+  const mockSessionManager = {
+    getEntries: vi.fn().mockReturnValue([]),
+    appendCustomEntry: vi.fn(),
+  } as unknown as SessionManager;
+
+  const mockMessages: AgentMessage[] = [{ role: "user", content: "hello" }];
+
+  beforeEach(() => {
+    vi.resetAllMocks();
+    vi.mocked(helpers.sanitizeSessionMessagesImages).mockImplementation(async (msgs) => msgs);
+    // Default mock implementation
+    vi.mocked(helpers.downgradeGeminiHistory).mockImplementation((msgs) => {
+      if (!msgs) return [];
+      return [...msgs, { role: "system", content: "downgraded" }];
+    });
+  });
+
+  it("should downgrade history for Google models if provider is not google-antigravity", async () => {
+    vi.mocked(helpers.isGoogleModelApi).mockReturnValue(true);
+
+    const result = await sanitizeSessionHistory({
+      messages: mockMessages,
+      modelApi: "google-gemini",
+      provider: "google-vertex",
+      sessionManager: mockSessionManager,
+      sessionId: "test-session",
+    });
+
+    expect(helpers.isGoogleModelApi).toHaveBeenCalledWith("google-gemini");
+    expect(helpers.downgradeGeminiHistory).toHaveBeenCalled();
+    // Check if the result contains the downgraded message
+    expect(result).toContainEqual({ role: "system", content: "downgraded" });
+  });
+
+  it("should NOT downgrade history for google-antigravity provider", async () => {
+    vi.mocked(helpers.isGoogleModelApi).mockReturnValue(true);
+
+    const result = await sanitizeSessionHistory({
+      messages: mockMessages,
+      modelApi: "google-gemini",
+      provider: "google-antigravity",
+      sessionManager: mockSessionManager,
+      sessionId: "test-session",
+    });
+
+    expect(helpers.isGoogleModelApi).toHaveBeenCalledWith("google-gemini");
+    expect(helpers.downgradeGeminiHistory).not.toHaveBeenCalled();
+    // Result should not contain the downgraded message
+    expect(result).not.toContainEqual({
+      role: "system",
+      content: "downgraded",
+    });
+  });
+
+  it("should NOT downgrade history for non-Google models", async () => {
+    vi.mocked(helpers.isGoogleModelApi).mockReturnValue(false);
+
+    const _result = await sanitizeSessionHistory({
+      messages: mockMessages,
+      modelApi: "anthropic-messages",
+      provider: "anthropic",
+      sessionManager: mockSessionManager,
+      sessionId: "test-session",
+    });
+
+    expect(helpers.isGoogleModelApi).toHaveBeenCalledWith("anthropic-messages");
+    expect(helpers.downgradeGeminiHistory).not.toHaveBeenCalled();
+  });
+
+  it("should downgrade history if provider is undefined but model is Google", async () => {
+    vi.mocked(helpers.isGoogleModelApi).mockReturnValue(true);
+
+    const _result = await sanitizeSessionHistory({
+      messages: mockMessages,
+      modelApi: "google-gemini",
+      provider: undefined,
+      sessionManager: mockSessionManager,
+      sessionId: "test-session",
+    });
+
+    expect(helpers.isGoogleModelApi).toHaveBeenCalledWith("google-gemini");
+    expect(helpers.downgradeGeminiHistory).toHaveBeenCalled();
+  });
+});
@@ -44,7 +44,11 @@ import {
 } from "../skills.js";
 import { filterBootstrapFilesForSession, loadWorkspaceBootstrapFiles } from "../workspace.js";
 import { buildEmbeddedExtensionPaths } from "./extensions.js";
-import { logToolSchemasForGoogle, sanitizeSessionHistory } from "./google.js";
+import {
+  logToolSchemasForGoogle,
+  sanitizeSessionHistory,
+  sanitizeToolsForGoogle,
+} from "./google.js";
 import { getDmHistoryLimitFromSessionKey, limitHistoryTurns } from "./history.js";
 import { resolveGlobalLane, resolveSessionLane } from "./lanes.js";
 import { log } from "./logger.js";
@@ -183,7 +187,7 @@ export async function compactEmbeddedPiSession(params: {
          warn: (message) => log.warn(`${message} (sessionKey=${sessionLabel})`),
        });
        const runAbortController = new AbortController();
-        const tools = createClawdbotCodingTools({
+        const toolsRaw = createClawdbotCodingTools({
          exec: {
            ...resolveExecToolDefaults(params.config),
            elevated: params.bashElevated,
@@ -200,6 +204,7 @@ export async function compactEmbeddedPiSession(params: {
          modelId,
          modelAuthMode: resolveModelAuthMode(model.provider, params.config),
        });
+        const tools = sanitizeToolsForGoogle({ tools: toolsRaw, provider });
        logToolSchemasForGoogle({ tools, provider });
        const machineName = await getMachineDisplayName();
        const runtimeChannel = normalizeMessageChannel(
@@ -303,6 +308,7 @@ export async function compactEmbeddedPiSession(params: {
              messages: session.messages,
              modelApi: model.api,
              modelId,
+              provider,
              sessionManager,
              sessionId: params.sessionId,
            });
@@ -0,0 +1,37 @@
+import { describe, expect, it } from "vitest";
+
+import type { AgentTool } from "@mariozechner/pi-agent-core";
+import { sanitizeToolsForGoogle } from "./google.js";
+
+describe("sanitizeToolsForGoogle", () => {
+  it("strips unsupported schema keywords for Google providers", () => {
+    const tool = {
+      name: "test",
+      description: "test",
+      parameters: {
+        type: "object",
+        additionalProperties: false,
+        properties: {
+          foo: {
+            type: "string",
+            format: "uuid",
+          },
+        },
+      },
+      execute: async () => ({ ok: true, content: [] }),
+    } as unknown as AgentTool;
+
+    const [sanitized] = sanitizeToolsForGoogle({
+      tools: [tool],
+      provider: "google-gemini-cli",
+    });
+
+    const params = sanitized.parameters as {
+      additionalProperties?: unknown;
+      properties?: Record<string, { format?: unknown }>;
+    };
+
+    expect(params.additionalProperties).toBeUndefined();
+    expect(params.properties?.foo?.format).toBeUndefined();
+  });
+});
@@ -1,4 +1,5 @@
 import type { AgentMessage, AgentTool } from "@mariozechner/pi-agent-core";
+import type { TSchema } from "@sinclair/typebox";
 import type { SessionManager } from "@mariozechner/pi-coding-agent";

 import { registerUnhandledRejectionHandler } from "../../infra/unhandled-rejections.js";
@@ -14,6 +15,7 @@ import { sanitizeToolUseResultPairing } from "../session-transcript-repair.js";
 import { log } from "./logger.js";
 import { describeUnknownError } from "./utils.js";
 import { isAntigravityClaude } from "../pi-embedded-helpers/google.js";
+import { cleanToolSchemaForGemini } from "../pi-tools.schema.js";

 const GOOGLE_TURN_ORDERING_CUSTOM_TYPE = "google-turn-ordering-bootstrap";
 const GOOGLE_SCHEMA_UNSUPPORTED_KEYWORDS = new Set([
@@ -80,17 +82,39 @@ function findUnsupportedSchemaKeywords(schema: unknown, path: string): string[]
  return violations;
 }

+export function sanitizeToolsForGoogle<
+  TSchemaType extends TSchema = TSchema,
+  TResult = unknown,
+>(params: {
+  tools: AgentTool<TSchemaType, TResult>[];
+  provider: string;
+}): AgentTool<TSchemaType, TResult>[] {
+  if (params.provider !== "google-antigravity" && params.provider !== "google-gemini-cli") {
+    return params.tools;
+  }
+  return params.tools.map((tool) => {
+    if (!tool.parameters || typeof tool.parameters !== "object") return tool;
+    return {
+      ...tool,
+      parameters: cleanToolSchemaForGemini(
+        tool.parameters as Record<string, unknown>,
+      ) as TSchemaType,
+    };
+  });
+}
+
 export function logToolSchemasForGoogle(params: { tools: AgentTool[]; provider: string }) {
  if (params.provider !== "google-antigravity" && params.provider !== "google-gemini-cli") {
    return;
  }
  const toolNames = params.tools.map((tool, index) => `${index}:${tool.name}`);
+  const tools = sanitizeToolsForGoogle(params);
  log.info("google tool schema snapshot", {
    provider: params.provider,
-    toolCount: params.tools.length,
+    toolCount: tools.length,
    tools: toolNames,
  });
-  for (const [index, tool] of params.tools.entries()) {
+  for (const [index, tool] of tools.entries()) {
    const violations = findUnsupportedSchemaKeywords(tool.parameters, `${tool.name}.parameters`);
    if (violations.length > 0) {
      log.warn("google tool schema has unsupported keywords", {
@@ -164,22 +188,34 @@ export async function sanitizeSessionHistory(params: {
  messages: AgentMessage[];
  modelApi?: string | null;
  modelId?: string;
+  provider?: string;
  sessionManager: SessionManager;
  sessionId: string;
 }): Promise<AgentMessage[]> {
  const isAntigravityClaudeModel = isAntigravityClaude(params.modelApi, params.modelId);
+  const provider = (params.provider ?? "").toLowerCase();
+  const modelId = (params.modelId ?? "").toLowerCase();
+  const isOpenRouterGemini =
+    (provider === "openrouter" || provider === "opencode") && modelId.includes("gemini");
+  const isGeminiLike = isGoogleModelApi(params.modelApi) || isOpenRouterGemini;
  const sanitizedImages = await sanitizeSessionMessagesImages(params.messages, "session:history", {
    sanitizeToolCallIds: shouldSanitizeToolCallIds(params.modelApi),
    enforceToolCallLast: params.modelApi === "anthropic-messages",
    preserveSignatures: params.modelApi === "google-antigravity" && isAntigravityClaudeModel,
+    sanitizeThoughtSignatures: isOpenRouterGemini
+      ? { allowBase64Only: true, includeCamelCase: true }
+      : undefined,
  });
  const repairedTools = sanitizeToolUseResultPairing(sanitizedImages);
-  const shouldDowngradeGemini = isGoogleModelApi(params.modelApi) && !isAntigravityClaudeModel;
+  const isAntigravityProvider =
+    provider === "google-antigravity" || params.modelApi === "google-antigravity";
+  const shouldDowngradeThinking = isGeminiLike && !isAntigravityClaudeModel;
  // Gemini rejects unsigned thinking blocks; downgrade them before send to avoid INVALID_ARGUMENT.
-  const downgradedThinking = shouldDowngradeGemini
+  const downgradedThinking = shouldDowngradeThinking
    ? downgradeGeminiThinkingBlocks(repairedTools)
    : repairedTools;
-  const downgraded = shouldDowngradeGemini
+  const shouldDowngradeHistory = shouldDowngradeThinking && !isAntigravityProvider;
+  const downgraded = shouldDowngradeHistory
    ? downgradeGeminiHistory(downgradedThinking)
    : downgradedThinking;

@@ -274,6 +274,8 @@ export async function runEmbeddedPiAgent(
                    provider,
                    model: model.id,
                  },
+                  systemPromptReport: attempt.systemPromptReport,
+                  error: { kind: "role_ordering", message: errorText },
                },
              };
            }
@@ -45,7 +45,11 @@ import { filterBootstrapFilesForSession, loadWorkspaceBootstrapFiles } from "../
 import { isAbortError } from "../abort.js";
 import { buildEmbeddedExtensionPaths } from "../extensions.js";
 import { applyExtraParamsToAgent } from "../extra-params.js";
-import { logToolSchemasForGoogle, sanitizeSessionHistory } from "../google.js";
+import {
+  logToolSchemasForGoogle,
+  sanitizeSessionHistory,
+  sanitizeToolsForGoogle,
+} from "../google.js";
 import { getDmHistoryLimitFromSessionKey, limitHistoryTurns } from "../history.js";
 import { log } from "../logger.js";
 import { buildModelAliasLines } from "../model.js";
@@ -127,7 +131,7 @@ export async function runEmbeddedAttempt(

    const agentDir = params.agentDir ?? resolveClawdbotAgentDir();

-    const tools = createClawdbotCodingTools({
+    const toolsRaw = createClawdbotCodingTools({
      exec: {
        ...resolveExecToolDefaults(params.config),
        elevated: params.bashElevated,
@@ -148,6 +152,7 @@ export async function runEmbeddedAttempt(
      replyToMode: params.replyToMode,
      hasRepliedRef: params.hasRepliedRef,
    });
+    const tools = sanitizeToolsForGoogle({ tools: toolsRaw, provider: params.provider });
    logToolSchemasForGoogle({ tools, provider: params.provider });

    const machineName = await getMachineDisplayName();
@@ -318,6 +323,7 @@ export async function runEmbeddedAttempt(
          messages: activeSession.messages,
          modelApi: params.model.api,
          modelId: params.modelId,
+          provider: params.provider,
          sessionManager,
          sessionId: params.sessionId,
        });
@@ -20,7 +20,7 @@ export type EmbeddedPiRunMeta = {
  aborted?: boolean;
  systemPromptReport?: SessionSystemPromptReport;
  error?: {
-    kind: "context_overflow" | "compaction_failure";
+    kind: "context_overflow" | "compaction_failure" | "role_ordering";
    message: string;
  };
 };
@@ -14,7 +14,7 @@ export function handleAgentStart(ctx: EmbeddedPiSubscribeContext) {
      startedAt: Date.now(),
    },
  });
-  ctx.params.onAgentEvent?.({
+  void ctx.params.onAgentEvent?.({
    stream: "lifecycle",
    data: { phase: "start" },
  });
@@ -24,7 +24,7 @@ export function handleAutoCompactionStart(ctx: EmbeddedPiSubscribeContext) {
  ctx.state.compactionInFlight = true;
  ctx.ensureCompactionPromise();
  ctx.log.debug(`embedded run compaction start: runId=${ctx.params.runId}`);
-  ctx.params.onAgentEvent?.({
+  void ctx.params.onAgentEvent?.({
    stream: "compaction",
    data: { phase: "start" },
  });
@@ -43,7 +43,7 @@ export function handleAutoCompactionEnd(
  } else {
    ctx.maybeResolveCompactionWait();
  }
-  ctx.params.onAgentEvent?.({
+  void ctx.params.onAgentEvent?.({
    stream: "compaction",
    data: { phase: "end", willRetry },
  });
@@ -59,7 +59,7 @@ export function handleAgentEnd(ctx: EmbeddedPiSubscribeContext) {
      endedAt: Date.now(),
    },
  });
-  ctx.params.onAgentEvent?.({
+  void ctx.params.onAgentEvent?.({
    stream: "lifecycle",
    data: { phase: "end" },
  });
@@ -118,7 +118,7 @@ export function handleMessageUpdate(
        mediaUrls: mediaUrls?.length ? mediaUrls : undefined,
      },
    });
-    ctx.params.onAgentEvent?.({
+    void ctx.params.onAgentEvent?.({
      stream: "assistant",
      data: {
        text: cleanedText,
@@ -11,7 +11,7 @@ import {
 } from "./pi-embedded-subscribe.tools.js";
 import { inferToolMetaFromArgs } from "./pi-embedded-utils.js";

-export function handleToolExecutionStart(
+export async function handleToolExecutionStart(
  ctx: EmbeddedPiSubscribeContext,
  evt: AgentEvent & { toolName: string; toolCallId: string; args: unknown },
 ) {
@@ -53,7 +53,8 @@ export function handleToolExecutionStart(
      args: args as Record<string, unknown>,
    },
  });
-  ctx.params.onAgentEvent?.({
+  // Await onAgentEvent to ensure typing indicator starts before tool summaries are emitted.
+  await ctx.params.onAgentEvent?.({
    stream: "tool",
    data: { phase: "start", name: toolName, toolCallId },
  });
@@ -108,7 +109,7 @@ export function handleToolExecutionUpdate(
      partialResult: sanitized,
    },
  });
-  ctx.params.onAgentEvent?.({
+  void ctx.params.onAgentEvent?.({
    stream: "tool",
    data: {
      phase: "update",
@@ -170,7 +171,7 @@ export function handleToolExecutionEnd(
      result: sanitizedResult,
    },
  });
-  ctx.params.onAgentEvent?.({
+  void ctx.params.onAgentEvent?.({
    stream: "tool",
    data: {
      phase: "result",
@@ -32,7 +32,11 @@ export function createEmbeddedPiSessionEventHandler(ctx: EmbeddedPiSubscribeCont
        handleMessageEnd(ctx, evt as never);
        return;
      case "tool_execution_start":
-        handleToolExecutionStart(ctx, evt as never);
+        // Async handler - awaits typing indicator before emitting tool summaries.
+        // Catch rejections to avoid unhandled promise rejection crashes.
+        handleToolExecutionStart(ctx, evt as never).catch((err) => {
+          ctx.log.debug(`tool_execution_start handler failed: ${String(err)}`);
+        });
        return;
      case "tool_execution_update":
        handleToolExecutionUpdate(ctx, evt as never);
@@ -13,7 +13,7 @@ describe("subscribeEmbeddedPiSession", () => {
    { tag: "antthinking", open: "<antthinking>", close: "</antthinking>" },
  ] as const;

-  it("includes canvas action metadata in tool summaries", () => {
+  it("includes canvas action metadata in tool summaries", async () => {
    let handler: ((evt: unknown) => void) | undefined;
    const session: StubSession = {
      subscribe: (fn) => {
@@ -38,6 +38,9 @@ describe("subscribeEmbeddedPiSession", () => {
      args: { action: "a2ui_push", jsonlPath: "/tmp/a2ui.jsonl" },
    });

+    // Wait for async handler to complete
+    await Promise.resolve();
+
    expect(onToolResult).toHaveBeenCalledTimes(1);
    const payload = onToolResult.mock.calls[0][0];
    expect(payload.text).toContain("🖼️");
@@ -72,7 +75,7 @@ describe("subscribeEmbeddedPiSession", () => {

    expect(onToolResult).not.toHaveBeenCalled();
  });
-  it("emits tool summaries when shouldEmitToolResult overrides verbose", () => {
+  it("emits tool summaries when shouldEmitToolResult overrides verbose", async () => {
    let handler: ((evt: unknown) => void) | undefined;
    const session: StubSession = {
      subscribe: (fn) => {
@@ -98,6 +101,9 @@ describe("subscribeEmbeddedPiSession", () => {
      args: { path: "/tmp/c.txt" },
    });

+    // Wait for async handler to complete
+    await Promise.resolve();
+
    expect(onToolResult).toHaveBeenCalledTimes(1);
  });
 });
@@ -14,7 +14,7 @@ describe("subscribeEmbeddedPiSession", () => {
    { tag: "antthinking", open: "<antthinking>", close: "</antthinking>" },
  ] as const;

-  it("suppresses message_end block replies when the message tool already sent", () => {
+  it("suppresses message_end block replies when the message tool already sent", async () => {
    let handler: ((evt: unknown) => void) | undefined;
    const session: StubSession = {
      subscribe: (fn) => {
@@ -41,6 +41,9 @@ describe("subscribeEmbeddedPiSession", () => {
      args: { action: "send", to: "+1555", message: messageText },
    });

+    // Wait for async handler to complete
+    await Promise.resolve();
+
    handler?.({
      type: "tool_execution_end",
      toolName: "message",
@@ -58,7 +61,7 @@ describe("subscribeEmbeddedPiSession", () => {

    expect(onBlockReply).not.toHaveBeenCalled();
  });
-  it("does not suppress message_end replies when message tool reports error", () => {
+  it("does not suppress message_end replies when message tool reports error", async () => {
    let handler: ((evt: unknown) => void) | undefined;
    const session: StubSession = {
      subscribe: (fn) => {
@@ -85,6 +88,9 @@ describe("subscribeEmbeddedPiSession", () => {
      args: { action: "send", to: "+1555", message: messageText },
    });

+    // Wait for async handler to complete
+    await Promise.resolve();
+
    handler?.({
      type: "tool_execution_end",
      toolName: "message",
@@ -54,7 +54,7 @@ describe("subscribeEmbeddedPiSession", () => {
    await waitPromise;
    expect(resolved).toBe(true);
  });
-  it("emits tool summaries at tool start when verbose is on", () => {
+  it("emits tool summaries at tool start when verbose is on", async () => {
    let handler: ((evt: unknown) => void) | undefined;
    const session: StubSession = {
      subscribe: (fn) => {
@@ -79,6 +79,9 @@ describe("subscribeEmbeddedPiSession", () => {
      args: { path: "/tmp/a.txt" },
    });

+    // Wait for async handler to complete
+    await Promise.resolve();
+
    expect(onToolResult).toHaveBeenCalledTimes(1);
    const payload = onToolResult.mock.calls[0][0];
    expect(payload.text).toContain("/tmp/a.txt");
@@ -93,7 +96,7 @@ describe("subscribeEmbeddedPiSession", () => {

    expect(onToolResult).toHaveBeenCalledTimes(1);
  });
-  it("includes browser action metadata in tool summaries", () => {
+  it("includes browser action metadata in tool summaries", async () => {
    let handler: ((evt: unknown) => void) | undefined;
    const session: StubSession = {
      subscribe: (fn) => {
@@ -118,6 +121,9 @@ describe("subscribeEmbeddedPiSession", () => {
      args: { action: "snapshot", targetUrl: "https://example.com" },
    });

+    // Wait for async handler to complete
+    await Promise.resolve();
+
    expect(onToolResult).toHaveBeenCalledTimes(1);
    const payload = onToolResult.mock.calls[0][0];
    expect(payload.text).toContain("🌐");
@@ -22,7 +22,7 @@ export type SubscribeEmbeddedPiSessionParams = {
  blockReplyChunking?: BlockReplyChunking;
  onPartialReply?: (payload: { text?: string; mediaUrls?: string[] }) => void | Promise<void>;
  onAssistantMessageStart?: () => void | Promise<void>;
-  onAgentEvent?: (evt: { stream: string; data: Record<string, unknown> }) => void;
+  onAgentEvent?: (evt: { stream: string; data: Record<string, unknown> }) => void | Promise<void>;
  enforceFinalTag?: boolean;
 };

@@ -51,6 +51,8 @@ function buildSandboxBrowserResolvedConfig(params: {
    cdpProtocol: "http",
    cdpHost,
    cdpIsLoopback: true,
+    remoteCdpTimeoutMs: 1500,
+    remoteCdpHandshakeTimeoutMs: 3000,
    color: DEFAULT_CLAWD_BROWSER_COLOR,
    executablePath: undefined,
    headless: params.headless,
@@ -102,7 +102,7 @@ export function ensureSkillsWatcher(params: { workspaceDir: string; config?: Cla
  if (!watchEnabled) {
    if (existing) {
      watchers.delete(workspaceDir);
-      existing.timer && clearTimeout(existing.timer);
+      if (existing.timer) clearTimeout(existing.timer);
      void existing.watcher.close().catch(() => {});
    }
    return;
@@ -115,7 +115,7 @@ export function ensureSkillsWatcher(params: { workspaceDir: string; config?: Cla
  }
  if (existing) {
    watchers.delete(workspaceDir);
-    existing.timer && clearTimeout(existing.timer);
+    if (existing.timer) clearTimeout(existing.timer);
    void existing.watcher.close().catch(() => {});
  }

@@ -173,7 +173,7 @@ export function buildAgentSystemPrompt(params: {
    browser: "Control web browser",
    canvas: "Present/eval/snapshot the Canvas",
    nodes: "List/describe/notify/camera/screen on paired nodes",
-    cron: "Manage cron jobs and wake events (use for reminders)",
+    cron: "Manage cron jobs and wake events (use for reminders; include recent context in reminder text if appropriate)",
    message: "Send messages and channel actions",
    gateway: "Restart, apply config, or run updates on the running Clawdbot process",
    agents_list: "List agent ids allowed for sessions_spawn",
@@ -319,7 +319,7 @@ export function buildAgentSystemPrompt(params: {
          "- browser: control clawd's dedicated browser",
          "- canvas: present/eval/snapshot the Canvas",
          "- nodes: list/describe/notify/camera/screen on paired nodes",
-          "- cron: manage cron jobs and wake events (use for reminders)",
+          "- cron: manage cron jobs and wake events (use for reminders; include recent context in reminder text if appropriate)",
          "- sessions_list: list sessions",
          "- sessions_history: fetch session history",
          "- sessions_send: send to another session",
@@ -84,4 +84,47 @@ describe("cron tool", () => {
      payload: { kind: "systemEvent", text: "hello" },
    });
  });
+
+  it("adds recent context for systemEvent reminders when session key is available", async () => {
+    callGatewayMock
+      .mockResolvedValueOnce({
+        messages: [
+          { role: "user", content: [{ type: "text", text: "Discussed Q2 budget" }] },
+          {
+            role: "assistant",
+            content: [{ type: "text", text: "We agreed to review on Tuesday." }],
+          },
+          { role: "user", content: [{ type: "text", text: "Remind me about the thing at 2pm" }] },
+        ],
+      })
+      .mockResolvedValueOnce({ ok: true });
+
+    const tool = createCronTool({ agentSessionKey: "main" });
+    await tool.execute("call3", {
+      action: "add",
+      job: {
+        name: "reminder",
+        schedule: { atMs: 123 },
+        payload: { text: "Reminder: the thing." },
+      },
+    });
+
+    expect(callGatewayMock).toHaveBeenCalledTimes(2);
+    const historyCall = callGatewayMock.mock.calls[0]?.[0] as {
+      method?: string;
+      params?: unknown;
+    };
+    expect(historyCall.method).toBe("chat.history");
+
+    const cronCall = callGatewayMock.mock.calls[1]?.[0] as {
+      method?: string;
+      params?: { payload?: { text?: string } };
+    };
+    expect(cronCall.method).toBe("cron.add");
+    const text = cronCall.params?.payload?.text ?? "";
+    expect(text).toContain("Recent context:");
+    expect(text).toContain("User: Discussed Q2 budget");
+    expect(text).toContain("Assistant: We agreed to review on Tuesday.");
+    expect(text).toContain("User: Remind me about the thing at 2pm");
+  });
 });
@@ -1,8 +1,11 @@
 import { Type } from "@sinclair/typebox";
 import { normalizeCronJobCreate, normalizeCronJobPatch } from "../../cron/normalize.js";
+import { loadConfig } from "../../config/config.js";
+import { truncateUtf16Safe } from "../../utils.js";
 import { optionalStringEnum, stringEnum } from "../schema/typebox.js";
 import { type AnyAgentTool, jsonResult, readStringParam } from "./common.js";
 import { callGatewayTool, type GatewayCallOptions } from "./gateway.js";
+import { resolveInternalSessionKey, resolveMainSessionAlias } from "./sessions-helpers.js";

 // NOTE: We use Type.Object({}, { additionalProperties: true }) for job/patch
 // instead of CronAddParamsSchema/CronJobPatchSchema because the gateway schemas
@@ -13,6 +16,11 @@ const CRON_ACTIONS = ["status", "list", "add", "update", "remove", "run", "runs"

 const CRON_WAKE_MODES = ["now", "next-heartbeat"] as const;

+const REMINDER_CONTEXT_MESSAGES = 3;
+const REMINDER_CONTEXT_PER_MESSAGE_MAX = 220;
+const REMINDER_CONTEXT_TOTAL_MAX = 700;
+const REMINDER_CONTEXT_MARKER = "\n\nRecent context:\n";
+
 // Flattened schema: runtime validates per-action requirements.
 const CronToolSchema = Type.Object({
  action: stringEnum(CRON_ACTIONS),
@@ -28,7 +36,90 @@ const CronToolSchema = Type.Object({
  mode: optionalStringEnum(CRON_WAKE_MODES),
 });

-export function createCronTool(): AnyAgentTool {
+type CronToolOptions = {
+  agentSessionKey?: string;
+};
+
+type ChatMessage = {
+  role?: unknown;
+  content?: unknown;
+};
+
+function stripExistingContext(text: string) {
+  const index = text.indexOf(REMINDER_CONTEXT_MARKER);
+  if (index === -1) return text;
+  return text.slice(0, index).trim();
+}
+
+function truncateText(input: string, maxLen: number) {
+  if (input.length <= maxLen) return input;
+  const truncated = truncateUtf16Safe(input, Math.max(0, maxLen - 3)).trimEnd();
+  return `${truncated}...`;
+}
+
+function normalizeContextText(raw: string) {
+  return raw.replace(/\s+/g, " ").trim();
+}
+
+function extractMessageText(message: ChatMessage): { role: string; text: string } | null {
+  const role = typeof message.role === "string" ? message.role : "";
+  if (role !== "user" && role !== "assistant") return null;
+  const content = message.content;
+  if (typeof content === "string") {
+    const normalized = normalizeContextText(content);
+    return normalized ? { role, text: normalized } : null;
+  }
+  if (!Array.isArray(content)) return null;
+  const chunks: string[] = [];
+  for (const block of content) {
+    if (!block || typeof block !== "object") continue;
+    if ((block as { type?: unknown }).type !== "text") continue;
+    const text = (block as { text?: unknown }).text;
+    if (typeof text === "string" && text.trim()) {
+      chunks.push(text);
+    }
+  }
+  const joined = normalizeContextText(chunks.join(" "));
+  return joined ? { role, text: joined } : null;
+}
+
+async function buildReminderContextLines(params: {
+  agentSessionKey?: string;
+  gatewayOpts: GatewayCallOptions;
+}) {
+  const sessionKey = params.agentSessionKey?.trim();
+  if (!sessionKey) return [];
+  const cfg = loadConfig();
+  const { mainKey, alias } = resolveMainSessionAlias(cfg);
+  const resolvedKey = resolveInternalSessionKey({ key: sessionKey, alias, mainKey });
+  try {
+    const res = (await callGatewayTool("chat.history", params.gatewayOpts, {
+      sessionKey: resolvedKey,
+      limit: 12,
+    })) as { messages?: unknown[] };
+    const messages = Array.isArray(res?.messages) ? res.messages : [];
+    const parsed = messages
+      .map((msg) => extractMessageText(msg as ChatMessage))
+      .filter((msg): msg is { role: string; text: string } => Boolean(msg));
+    const recent = parsed.slice(-REMINDER_CONTEXT_MESSAGES);
+    if (recent.length === 0) return [];
+    const lines: string[] = [];
+    let total = 0;
+    for (const entry of recent) {
+      const label = entry.role === "user" ? "User" : "Assistant";
+      const text = truncateText(entry.text, REMINDER_CONTEXT_PER_MESSAGE_MAX);
+      const line = `- ${label}: ${text}`;
+      total += line.length;
+      if (total > REMINDER_CONTEXT_TOTAL_MAX) break;
+      lines.push(line);
+    }
+    return lines;
+  } catch {
+    return [];
+  }
+}
+
+export function createCronTool(opts?: CronToolOptions): AnyAgentTool {
  return {
    label: "Cron",
    name: "cron",
@@ -58,6 +149,24 @@ export function createCronTool(): AnyAgentTool {
            throw new Error("job required");
          }
          const job = normalizeCronJobCreate(params.job) ?? params.job;
+          if (
+            job &&
+            typeof job === "object" &&
+            "payload" in job &&
+            (job as { payload?: { kind?: string; text?: string } }).payload?.kind === "systemEvent"
+          ) {
+            const payload = (job as { payload: { kind: string; text: string } }).payload;
+            if (typeof payload.text === "string" && payload.text.trim()) {
+              const contextLines = await buildReminderContextLines({
+                agentSessionKey: opts?.agentSessionKey,
+                gatewayOpts,
+              });
+              if (contextLines.length > 0) {
+                const baseText = stripExistingContext(payload.text);
+                payload.text = `${baseText}${REMINDER_CONTEXT_MARKER}${contextLines.join("\n")}`;
+              }
+            }
+          }
          return jsonResult(await callGatewayTool("cron.add", gatewayOpts, job));
        }
        case "update": {
@@ -219,7 +219,7 @@ export async function handleDiscordGuildAction(
      return jsonResult({ ok: true, event });
    }
    case "channelCreate": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const guildId = readStringParam(params, "guildId", { required: true });
@@ -241,7 +241,7 @@ export async function handleDiscordGuildAction(
      return jsonResult({ ok: true, channel });
    }
    case "channelEdit": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const channelId = readStringParam(params, "channelId", {
@@ -267,7 +267,7 @@ export async function handleDiscordGuildAction(
      return jsonResult({ ok: true, channel });
    }
    case "channelDelete": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const channelId = readStringParam(params, "channelId", {
@@ -277,7 +277,7 @@ export async function handleDiscordGuildAction(
      return jsonResult(result);
    }
    case "channelMove": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const guildId = readStringParam(params, "guildId", { required: true });
@@ -295,7 +295,7 @@ export async function handleDiscordGuildAction(
      return jsonResult({ ok: true });
    }
    case "categoryCreate": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const guildId = readStringParam(params, "guildId", { required: true });
@@ -310,7 +310,7 @@ export async function handleDiscordGuildAction(
      return jsonResult({ ok: true, category: channel });
    }
    case "categoryEdit": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const categoryId = readStringParam(params, "categoryId", {
@@ -326,7 +326,7 @@ export async function handleDiscordGuildAction(
      return jsonResult({ ok: true, category: channel });
    }
    case "categoryDelete": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const categoryId = readStringParam(params, "categoryId", {
@@ -336,7 +336,7 @@ export async function handleDiscordGuildAction(
      return jsonResult(result);
    }
    case "channelPermissionSet": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const channelId = readStringParam(params, "channelId", {
@@ -359,7 +359,7 @@ export async function handleDiscordGuildAction(
      return jsonResult({ ok: true });
    }
    case "channelPermissionRemove": {
-      if (!isActionEnabled("channels", false)) {
+      if (!isActionEnabled("channels")) {
        throw new Error("Discord channel management is disabled.");
      }
      const channelId = readStringParam(params, "channelId", {
@@ -1,6 +1,5 @@
 import { Type } from "@sinclair/typebox";
 import { resolveAgentDir } from "../../agents/agent-scope.js";
-import { buildAuthHealthSummary, formatRemainingShort } from "../../agents/auth-health.js";
 import {
  ensureAuthProfileStore,
  resolveAuthProfileDisplayLabel,
@@ -32,7 +31,6 @@ import {
  formatUsageWindowSummary,
  loadProviderUsageSummary,
  resolveUsageProviderId,
-  type UsageProviderId,
 } from "../../infra/provider-usage.js";
 import {
  buildAgentMainSessionKey,
@@ -283,48 +281,33 @@ export function createSessionStatusTool(opts?: {
        defaultModel: DEFAULT_MODEL,
      });
      const providerForCard = resolved.entry.providerOverride?.trim() || configured.provider;
-      const authStore = ensureAuthProfileStore(agentDir, { allowKeychainPrompt: false });
-      const authHealth = buildAuthHealthSummary({
-        store: authStore,
-        cfg,
-      });
-      const oauthProfiles = authHealth.profiles.filter(
-        (profile) => profile.type === "oauth" || profile.type === "token",
-      );
-
-      const usageProviders = Array.from(
-        new Set(
-          oauthProfiles
-            .map((profile) => resolveUsageProviderId(profile.provider))
-            .filter((provider): provider is UsageProviderId => Boolean(provider)),
-        ),
-      );
-      const usageByProvider = new Map<string, string>();
-      if (usageProviders.length > 0) {
+      const usageProvider = resolveUsageProviderId(providerForCard);
+      let usageLine: string | undefined;
+      if (usageProvider) {
        try {
          const usageSummary = await loadProviderUsageSummary({
            timeoutMs: 3500,
-            providers: usageProviders,
+            providers: [usageProvider],
            agentDir,
          });
-          for (const snapshot of usageSummary.providers) {
+          const snapshot = usageSummary.providers.find(
+            (entry) => entry.provider === usageProvider,
+          );
+          if (snapshot) {
            const formatted = formatUsageWindowSummary(snapshot, {
              now: Date.now(),
              maxWindows: 2,
+              includeResets: true,
            });
-            if (formatted) usageByProvider.set(snapshot.provider, formatted);
+            if (formatted && !formatted.startsWith("error:")) {
+              usageLine = `📊 Usage: ${formatted}`;
+            }
          }
        } catch {
          // ignore
        }
      }

-      const usageProvider = resolveUsageProviderId(providerForCard);
-      const usageLine =
-        oauthProfiles.length === 0 && usageProvider && usageByProvider.has(usageProvider)
-          ? `📊 Usage: ${usageByProvider.get(usageProvider)}`
-          : undefined;
-
      const isGroup =
        resolved.entry.chatType === "group" ||
        resolved.entry.chatType === "room" ||
@@ -370,53 +353,13 @@ export function createSessionStatusTool(opts?: {
        includeTranscriptUsage: false,
      });

-      const authStatusLines = (() => {
-        if (oauthProfiles.length === 0) return [];
-        const formatStatus = (status: string) => {
-          if (status === "ok") return "ok";
-          if (status === "static") return "static";
-          if (status === "expiring") return "expiring";
-          if (status === "missing") return "unknown";
-          return "expired";
-        };
-        const profilesByProvider = new Map<string, typeof oauthProfiles>();
-        for (const profile of oauthProfiles) {
-          const current = profilesByProvider.get(profile.provider);
-          if (current) current.push(profile);
-          else profilesByProvider.set(profile.provider, [profile]);
-        }
-        const lines: string[] = ["OAuth/token status"];
-        for (const [provider, profiles] of profilesByProvider) {
-          const usageKey = resolveUsageProviderId(provider);
-          const usage = usageKey ? usageByProvider.get(usageKey) : undefined;
-          const usageSuffix = usage ? ` — usage: ${usage}` : "";
-          lines.push(`- ${provider}${usageSuffix}`);
-          for (const profile of profiles) {
-            const labelText = profile.label || profile.profileId;
-            const status = formatStatus(profile.status);
-            const expiry =
-              profile.status === "static"
-                ? ""
-                : profile.expiresAt
-                  ? ` expires in ${formatRemainingShort(profile.remainingMs)}`
-                  : " expires unknown";
-            const source = profile.source !== "store" ? ` (${profile.source})` : "";
-            lines.push(`  - ${labelText} ${status}${expiry}${source}`);
-          }
-        }
-        return lines;
-      })();
-
-      const fullStatusText =
-        authStatusLines.length > 0 ? `${statusText}\n\n${authStatusLines.join("\n")}` : statusText;
-
      return {
-        content: [{ type: "text", text: fullStatusText }],
+        content: [{ type: "text", text: statusText }],
        details: {
          ok: true,
          sessionKey: resolved.key,
          changedModel,
-          statusText: fullStatusText,
+          statusText,
        },
      };
    },
@@ -26,7 +26,7 @@ import type { VerboseLevel } from "../thinking.js";
 import { isSilentReplyText, SILENT_REPLY_TOKEN } from "../tokens.js";
 import type { GetReplyOptions, ReplyPayload } from "../types.js";
 import { buildThreadingToolContext, resolveEnforceFinalTag } from "./agent-runner-utils.js";
-import type { BlockReplyPipeline } from "./block-reply-pipeline.js";
+import { createBlockReplyPayloadKey, type BlockReplyPipeline } from "./block-reply-pipeline.js";
 import type { FollowupRun } from "./queue.js";
 import { parseReplyDirectives } from "./reply-directives.js";
 import { applyReplyTagsToPayload, isRenderablePayload } from "./reply-payloads.js";
@@ -40,6 +40,8 @@ export type AgentRunLoopResult =
      fallbackModel?: string;
      didLogHeartbeatStrip: boolean;
      autoCompactionCompleted: boolean;
+      /** Payload keys sent directly (not via pipeline) during tool flush. */
+      directlySentBlockKeys?: Set<string>;
    }
  | { kind: "final"; payload: ReplyPayload };

@@ -61,6 +63,7 @@ export async function runAgentTurnWithFallback(params: {
  shouldEmitToolResult: () => boolean;
  pendingToolTasks: Set<Promise<void>>;
  resetSessionAfterCompactionFailure: (reason: string) => Promise<boolean>;
+  resetSessionAfterRoleOrderingConflict: (reason: string) => Promise<boolean>;
  isHeartbeat: boolean;
  sessionKey?: string;
  getActiveSessionEntry: () => SessionEntry | undefined;
@@ -70,6 +73,8 @@ export async function runAgentTurnWithFallback(params: {
 }): Promise<AgentRunLoopResult> {
  let didLogHeartbeatStrip = false;
  let autoCompactionCompleted = false;
+  // Track payloads sent directly (not via pipeline) during tool flush to avoid duplicates.
+  const directlySentBlockKeys = new Set<string>();

  const runId = crypto.randomUUID();
  if (params.sessionKey) {
@@ -244,12 +249,13 @@ export async function runAgentTurnWithFallback(params: {
                    });
                  }
                : undefined,
-            onAgentEvent: (evt) => {
-              // Trigger typing when tools start executing
+            onAgentEvent: async (evt) => {
+              // Trigger typing when tools start executing.
+              // Must await to ensure typing indicator starts before tool summaries are emitted.
              if (evt.stream === "tool") {
                const phase = typeof evt.data.phase === "string" ? evt.data.phase : "";
                if (phase === "start" || phase === "update") {
-                  void params.typingSignals.signalToolStart();
+                  await params.typingSignals.signalToolStart();
                }
              }
              // Track auto-compaction completion
@@ -261,57 +267,67 @@ export async function runAgentTurnWithFallback(params: {
                }
              }
            },
-            onBlockReply:
-              params.blockStreamingEnabled && params.opts?.onBlockReply
-                ? async (payload) => {
-                    const { text, skip } = normalizeStreamingText(payload);
-                    const hasPayloadMedia = (payload.mediaUrls?.length ?? 0) > 0;
-                    if (skip && !hasPayloadMedia) return;
-                    const taggedPayload = applyReplyTagsToPayload(
-                      {
-                        text,
-                        mediaUrls: payload.mediaUrls,
-                        mediaUrl: payload.mediaUrls?.[0],
-                      },
-                      params.sessionCtx.MessageSid,
-                    );
-                    // Let through payloads with audioAsVoice flag even if empty (need to track it)
-                    if (!isRenderablePayload(taggedPayload) && !payload.audioAsVoice) return;
-                    const parsed = parseReplyDirectives(taggedPayload.text ?? "", {
-                      currentMessageId: params.sessionCtx.MessageSid,
-                      silentToken: SILENT_REPLY_TOKEN,
-                    });
-                    const cleaned = parsed.text || undefined;
-                    const hasRenderableMedia =
-                      Boolean(taggedPayload.mediaUrl) || (taggedPayload.mediaUrls?.length ?? 0) > 0;
-                    // Skip empty payloads unless they have audioAsVoice flag (need to track it)
-                    if (
-                      !cleaned &&
-                      !hasRenderableMedia &&
-                      !payload.audioAsVoice &&
-                      !parsed.audioAsVoice
-                    )
-                      return;
-                    if (parsed.isSilent && !hasRenderableMedia) return;
+            // Always pass onBlockReply so flushBlockReplyBuffer works before tool execution,
+            // even when regular block streaming is disabled. The handler sends directly
+            // via opts.onBlockReply when the pipeline isn't available.
+            onBlockReply: params.opts?.onBlockReply
+              ? async (payload) => {
+                  const { text, skip } = normalizeStreamingText(payload);
+                  const hasPayloadMedia = (payload.mediaUrls?.length ?? 0) > 0;
+                  if (skip && !hasPayloadMedia) return;
+                  const taggedPayload = applyReplyTagsToPayload(
+                    {
+                      text,
+                      mediaUrls: payload.mediaUrls,
+                      mediaUrl: payload.mediaUrls?.[0],
+                    },
+                    params.sessionCtx.MessageSid,
+                  );
+                  // Let through payloads with audioAsVoice flag even if empty (need to track it)
+                  if (!isRenderablePayload(taggedPayload) && !payload.audioAsVoice) return;
+                  const parsed = parseReplyDirectives(taggedPayload.text ?? "", {
+                    currentMessageId: params.sessionCtx.MessageSid,
+                    silentToken: SILENT_REPLY_TOKEN,
+                  });
+                  const cleaned = parsed.text || undefined;
+                  const hasRenderableMedia =
+                    Boolean(taggedPayload.mediaUrl) || (taggedPayload.mediaUrls?.length ?? 0) > 0;
+                  // Skip empty payloads unless they have audioAsVoice flag (need to track it)
+                  if (
+                    !cleaned &&
+                    !hasRenderableMedia &&
+                    !payload.audioAsVoice &&
+                    !parsed.audioAsVoice
+                  )
+                    return;
+                  if (parsed.isSilent && !hasRenderableMedia) return;

-                    const blockPayload: ReplyPayload = params.applyReplyToMode({
-                      ...taggedPayload,
-                      text: cleaned,
-                      audioAsVoice: Boolean(parsed.audioAsVoice || payload.audioAsVoice),
-                      replyToId: taggedPayload.replyToId ?? parsed.replyToId,
-                      replyToTag: taggedPayload.replyToTag || parsed.replyToTag,
-                      replyToCurrent: taggedPayload.replyToCurrent || parsed.replyToCurrent,
+                  const blockPayload: ReplyPayload = params.applyReplyToMode({
+                    ...taggedPayload,
+                    text: cleaned,
+                    audioAsVoice: Boolean(parsed.audioAsVoice || payload.audioAsVoice),
+                    replyToId: taggedPayload.replyToId ?? parsed.replyToId,
+                    replyToTag: taggedPayload.replyToTag || parsed.replyToTag,
+                    replyToCurrent: taggedPayload.replyToCurrent || parsed.replyToCurrent,
+                  });
+
+                  void params.typingSignals
+                    .signalTextDelta(cleaned ?? taggedPayload.text)
+                    .catch((err) => {
+                      logVerbose(`block reply typing signal failed: ${String(err)}`);
                    });

-                    void params.typingSignals
-                      .signalTextDelta(cleaned ?? taggedPayload.text)
-                      .catch((err) => {
-                        logVerbose(`block reply typing signal failed: ${String(err)}`);
-                      });
-
-                    params.blockReplyPipeline?.enqueue(blockPayload);
+                  // Use pipeline if available (block streaming enabled), otherwise send directly
+                  if (params.blockStreamingEnabled && params.blockReplyPipeline) {
+                    params.blockReplyPipeline.enqueue(blockPayload);
+                  } else {
+                    // Send directly when flushing before tool execution (no streaming).
+                    // Track sent key to avoid duplicate in final payloads.
+                    directlySentBlockKeys.add(createBlockReplyPayloadKey(blockPayload));
+                    await params.opts?.onBlockReply?.(blockPayload);
                  }
-                : undefined,
+                }
+              : undefined,
            onBlockReplyFlush:
              params.blockStreamingEnabled && blockReplyPipeline
                ? async () => {
@@ -361,6 +377,17 @@ export async function runAgentTurnWithFallback(params: {
        didResetAfterCompactionFailure = true;
        continue;
      }
+      if (embeddedError?.kind === "role_ordering") {
+        const didReset = await params.resetSessionAfterRoleOrderingConflict(embeddedError.message);
+        if (didReset) {
+          return {
+            kind: "final",
+            payload: {
+              text: "⚠️ Message ordering conflict. I've reset the conversation - please try again.",
+            },
+          };
+        }
+      }

      break;
    } catch (err) {
@@ -380,6 +407,17 @@ export async function runAgentTurnWithFallback(params: {
        didResetAfterCompactionFailure = true;
        continue;
      }
+      if (isRoleOrderingError) {
+        const didReset = await params.resetSessionAfterRoleOrderingConflict(message);
+        if (didReset) {
+          return {
+            kind: "final",
+            payload: {
+              text: "⚠️ Message ordering conflict. I've reset the conversation - please try again.",
+            },
+          };
+        }
+      }

      // Auto-recover from Gemini session corruption by resetting the session
      if (
@@ -447,5 +485,6 @@ export async function runAgentTurnWithFallback(params: {
    fallbackModel,
    didLogHeartbeatStrip,
    autoCompactionCompleted,
+    directlySentBlockKeys: directlySentBlockKeys.size > 0 ? directlySentBlockKeys : undefined,
  };
 }
@@ -5,7 +5,7 @@ import type { OriginatingChannelType } from "../templating.js";
 import { SILENT_REPLY_TOKEN } from "../tokens.js";
 import type { ReplyPayload } from "../types.js";
 import { formatBunFetchSocketError, isBunFetchSocketError } from "./agent-runner-utils.js";
-import type { BlockReplyPipeline } from "./block-reply-pipeline.js";
+import { createBlockReplyPayloadKey, type BlockReplyPipeline } from "./block-reply-pipeline.js";
 import { parseReplyDirectives } from "./reply-directives.js";
 import {
  applyReplyThreading,
@@ -20,6 +20,8 @@ export function buildReplyPayloads(params: {
  didLogHeartbeatStrip: boolean;
  blockStreamingEnabled: boolean;
  blockReplyPipeline: BlockReplyPipeline | null;
+  /** Payload keys sent directly (not via pipeline) during tool flush. */
+  directlySentBlockKeys?: Set<string>;
  replyToMode: ReplyToMode;
  replyToChannel?: OriginatingChannelType;
  currentMessageId?: string;
@@ -98,11 +100,16 @@ export function buildReplyPayloads(params: {
    payloads: replyTaggedPayloads,
    sentTexts: messagingToolSentTexts,
  });
+  // Filter out payloads already sent via pipeline or directly during tool flush.
  const filteredPayloads = shouldDropFinalPayloads
    ? []
    : params.blockStreamingEnabled
      ? dedupedPayloads.filter((payload) => !params.blockReplyPipeline?.hasSentPayload(payload))
-      : dedupedPayloads;
+      : params.directlySentBlockKeys?.size
+        ? dedupedPayloads.filter(
+            (payload) => !params.directlySentBlockKeys!.has(createBlockReplyPayloadKey(payload)),
+          )
+        : dedupedPayloads;
  const replyPayloads = suppressMessagingToolReplies ? [] : filteredPayloads;

  return {
@@ -3,6 +3,7 @@ import { tmpdir } from "node:os";
 import path from "node:path";
 import { describe, expect, it, vi } from "vitest";
 import type { SessionEntry } from "../../config/sessions.js";
+import * as sessions from "../../config/sessions.js";
 import type { TypingMode } from "../../config/types.js";
 import type { TemplateContext } from "../templating.js";
 import type { GetReplyOptions } from "../types.js";
@@ -127,11 +128,14 @@ describe("runReplyAgent typing (heartbeat)", () => {
    try {
      const sessionId = "session";
      const storePath = path.join(stateDir, "sessions", "sessions.json");
-      const sessionEntry = { sessionId, updatedAt: Date.now() };
+      const transcriptPath = sessions.resolveSessionTranscriptPath(sessionId);
+      const sessionEntry = { sessionId, updatedAt: Date.now(), sessionFile: transcriptPath };
      const sessionStore = { main: sessionEntry };

      await fs.mkdir(path.dirname(storePath), { recursive: true });
      await fs.writeFile(storePath, JSON.stringify(sessionStore), "utf-8");
+      await fs.mkdir(path.dirname(transcriptPath), { recursive: true });
+      await fs.writeFile(transcriptPath, "ok", "utf-8");

      runEmbeddedPiAgentMock
        .mockImplementationOnce(async () => {
@@ -175,11 +179,14 @@ describe("runReplyAgent typing (heartbeat)", () => {
    try {
      const sessionId = "session";
      const storePath = path.join(stateDir, "sessions", "sessions.json");
-      const sessionEntry = { sessionId, updatedAt: Date.now() };
+      const transcriptPath = sessions.resolveSessionTranscriptPath(sessionId);
+      const sessionEntry = { sessionId, updatedAt: Date.now(), sessionFile: transcriptPath };
      const sessionStore = { main: sessionEntry };

      await fs.mkdir(path.dirname(storePath), { recursive: true });
      await fs.writeFile(storePath, JSON.stringify(sessionStore), "utf-8");
+      await fs.mkdir(path.dirname(transcriptPath), { recursive: true });
+      await fs.writeFile(transcriptPath, "ok", "utf-8");

      runEmbeddedPiAgentMock
        .mockImplementationOnce(async () => ({
@@ -212,6 +219,59 @@ describe("runReplyAgent typing (heartbeat)", () => {
      expect(payload).toMatchObject({ text: "ok" });
      expect(sessionStore.main.sessionId).not.toBe(sessionId);

+      const persisted = JSON.parse(await fs.readFile(storePath, "utf-8"));
+      expect(persisted.main.sessionId).toBe(sessionStore.main.sessionId);
+    } finally {
+      if (prevStateDir) {
+        process.env.CLAWDBOT_STATE_DIR = prevStateDir;
+      } else {
+        delete process.env.CLAWDBOT_STATE_DIR;
+      }
+    }
+  });
+  it("resets the session after role ordering payloads", async () => {
+    const prevStateDir = process.env.CLAWDBOT_STATE_DIR;
+    const stateDir = await fs.mkdtemp(path.join(tmpdir(), "clawdbot-session-role-ordering-"));
+    process.env.CLAWDBOT_STATE_DIR = stateDir;
+    try {
+      const sessionId = "session";
+      const storePath = path.join(stateDir, "sessions", "sessions.json");
+      const transcriptPath = sessions.resolveSessionTranscriptPath(sessionId);
+      const sessionEntry = { sessionId, updatedAt: Date.now(), sessionFile: transcriptPath };
+      const sessionStore = { main: sessionEntry };
+
+      await fs.mkdir(path.dirname(storePath), { recursive: true });
+      await fs.writeFile(storePath, JSON.stringify(sessionStore), "utf-8");
+      await fs.mkdir(path.dirname(transcriptPath), { recursive: true });
+      await fs.writeFile(transcriptPath, "ok", "utf-8");
+
+      runEmbeddedPiAgentMock.mockImplementationOnce(async () => ({
+        payloads: [{ text: "Message ordering conflict - please try again.", isError: true }],
+        meta: {
+          durationMs: 1,
+          error: {
+            kind: "role_ordering",
+            message: 'messages: roles must alternate between "user" and "assistant"',
+          },
+        },
+      }));
+
+      const { run } = createMinimalRun({
+        sessionEntry,
+        sessionStore,
+        sessionKey: "main",
+        storePath,
+      });
+      const res = await run();
+
+      const payload = Array.isArray(res) ? res[0] : res;
+      expect(payload).toMatchObject({
+        text: expect.stringContaining("Message ordering conflict"),
+      });
+      expect(payload.text?.toLowerCase()).toContain("reset");
+      expect(sessionStore.main.sessionId).not.toBe(sessionId);
+      await expect(fs.access(transcriptPath)).rejects.toBeDefined();
+
      const persisted = JSON.parse(await fs.readFile(storePath, "utf-8"));
      expect(persisted.main.sessionId).toBe(sessionStore.main.sessionId);
    } finally {
@@ -1,4 +1,5 @@
 import crypto from "node:crypto";
+import fs from "node:fs";
 import { setCliSessionId } from "../../agents/cli-session.js";
 import { lookupContextTokens } from "../../agents/context.js";
 import { DEFAULT_CONTEXT_TOKENS } from "../../agents/defaults.js";
@@ -8,6 +9,7 @@ import { queueEmbeddedPiMessage } from "../../agents/pi-embedded.js";
 import { hasNonzeroUsage } from "../../agents/usage.js";
 import {
  resolveAgentIdFromSessionKey,
+  resolveSessionFilePath,
  resolveSessionTranscriptPath,
  type SessionEntry,
  updateSessionStore,
@@ -209,11 +211,23 @@ export async function runReplyAgent(params: {
  });

  let responseUsageLine: string | undefined;
-  const resetSessionAfterCompactionFailure = async (reason: string): Promise<boolean> => {
+  type SessionResetOptions = {
+    failureLabel: string;
+    buildLogMessage: (nextSessionId: string) => string;
+    cleanupTranscripts?: boolean;
+  };
+  const resetSession = async ({
+    failureLabel,
+    buildLogMessage,
+    cleanupTranscripts,
+  }: SessionResetOptions): Promise<boolean> => {
    if (!sessionKey || !activeSessionStore || !storePath) return false;
+    const prevEntry = activeSessionStore[sessionKey] ?? activeSessionEntry;
+    if (!prevEntry) return false;
+    const prevSessionId = cleanupTranscripts ? prevEntry.sessionId : undefined;
    const nextSessionId = crypto.randomUUID();
    const nextEntry: SessionEntry = {
-      ...(activeSessionStore[sessionKey] ?? activeSessionEntry),
+      ...prevEntry,
      sessionId: nextSessionId,
      updatedAt: Date.now(),
      systemSent: false,
@@ -233,18 +247,42 @@ export async function runReplyAgent(params: {
      });
    } catch (err) {
      defaultRuntime.error(
-        `Failed to persist session reset after compaction failure (${sessionKey}): ${String(err)}`,
+        `Failed to persist session reset after ${failureLabel} (${sessionKey}): ${String(err)}`,
      );
    }
    followupRun.run.sessionId = nextSessionId;
    followupRun.run.sessionFile = nextSessionFile;
    activeSessionEntry = nextEntry;
    activeIsNewSession = true;
-    defaultRuntime.error(
-      `Auto-compaction failed (${reason}). Restarting session ${sessionKey} -> ${nextSessionId} and retrying.`,
-    );
+    defaultRuntime.error(buildLogMessage(nextSessionId));
+    if (cleanupTranscripts && prevSessionId) {
+      const transcriptCandidates = new Set<string>();
+      const resolved = resolveSessionFilePath(prevSessionId, prevEntry, { agentId });
+      if (resolved) transcriptCandidates.add(resolved);
+      transcriptCandidates.add(resolveSessionTranscriptPath(prevSessionId, agentId));
+      for (const candidate of transcriptCandidates) {
+        try {
+          fs.unlinkSync(candidate);
+        } catch {
+          // Best-effort cleanup.
+        }
+      }
+    }
    return true;
  };
+  const resetSessionAfterCompactionFailure = async (reason: string): Promise<boolean> =>
+    resetSession({
+      failureLabel: "compaction failure",
+      buildLogMessage: (nextSessionId) =>
+        `Auto-compaction failed (${reason}). Restarting session ${sessionKey} -> ${nextSessionId} and retrying.`,
+    });
+  const resetSessionAfterRoleOrderingConflict = async (reason: string): Promise<boolean> =>
+    resetSession({
+      failureLabel: "role ordering conflict",
+      buildLogMessage: (nextSessionId) =>
+        `Role ordering conflict (${reason}). Restarting session ${sessionKey} -> ${nextSessionId}.`,
+      cleanupTranscripts: true,
+    });
  try {
    const runOutcome = await runAgentTurnWithFallback({
      commandBody,
@@ -260,6 +298,7 @@ export async function runReplyAgent(params: {
      shouldEmitToolResult,
      pendingToolTasks,
      resetSessionAfterCompactionFailure,
+      resetSessionAfterRoleOrderingConflict,
      isHeartbeat,
      sessionKey,
      getActiveSessionEntry: () => activeSessionEntry,
@@ -272,7 +311,7 @@ export async function runReplyAgent(params: {
      return finalizeWithFollowup(runOutcome.payload, queueKey, runFollowupTurn);
    }

-    const { runResult, fallbackProvider, fallbackModel } = runOutcome;
+    const { runResult, fallbackProvider, fallbackModel, directlySentBlockKeys } = runOutcome;
    let { didLogHeartbeatStrip, autoCompactionCompleted } = runOutcome;

    if (
@@ -314,6 +353,7 @@ export async function runReplyAgent(params: {
      didLogHeartbeatStrip,
      blockStreamingEnabled,
      blockReplyPipeline,
+      directlySentBlockKeys,
      replyToMode,
      replyToChannel,
      currentMessageId: sessionCtx.MessageSid,
@@ -168,6 +168,7 @@ export async function buildStatusReply(params: {
        const formatted = formatUsageWindowSummary(snapshot, {
          now: Date.now(),
          maxWindows: 2,
+          includeResets: true,
        });
        if (formatted) usageByProvider.set(snapshot.provider, formatted);
      }
@@ -0,0 +1,27 @@
+import { describe, expect, it } from "vitest";
+
+import { appendCdpPath, getHeadersWithAuth } from "./cdp.helpers.js";
+
+describe("cdp.helpers", () => {
+  it("preserves query params when appending CDP paths", () => {
+    const url = appendCdpPath("https://example.com?token=abc", "/json/version");
+    expect(url).toBe("https://example.com/json/version?token=abc");
+  });
+
+  it("appends paths under a base prefix", () => {
+    const url = appendCdpPath("https://example.com/chrome/?token=abc", "json/list");
+    expect(url).toBe("https://example.com/chrome/json/list?token=abc");
+  });
+
+  it("adds basic auth headers when credentials are present", () => {
+    const headers = getHeadersWithAuth("https://user:pass@example.com");
+    expect(headers.Authorization).toBe(`Basic ${Buffer.from("user:pass").toString("base64")}`);
+  });
+
+  it("keeps preexisting authorization headers", () => {
+    const headers = getHeadersWithAuth("https://user:pass@example.com", {
+      Authorization: "Bearer token",
+    });
+    expect(headers.Authorization).toBe("Bearer token");
+  });
+});
@@ -28,6 +28,29 @@ export function isLoopbackHost(host: string) {
  );
 }

+export function getHeadersWithAuth(url: string, headers: Record<string, string> = {}) {
+  try {
+    const parsed = new URL(url);
+    const hasAuthHeader = Object.keys(headers).some((key) => key.toLowerCase() === "authorization");
+    if (hasAuthHeader) return headers;
+    if (parsed.username || parsed.password) {
+      const auth = Buffer.from(`${parsed.username}:${parsed.password}`).toString("base64");
+      return { ...headers, Authorization: `Basic ${auth}` };
+    }
+  } catch {
+    // ignore
+  }
+  return headers;
+}
+
+export function appendCdpPath(cdpUrl: string, path: string): string {
+  const url = new URL(cdpUrl);
+  const basePath = url.pathname.replace(/\/$/, "");
+  const suffix = path.startsWith("/") ? path : `/${path}`;
+  url.pathname = `${basePath}${suffix}`;
+  return url.toString();
+}
+
 function createCdpSender(ws: WebSocket) {
  let nextId = 1;
  const pending = new Map<number, Pending>();
@@ -75,11 +98,12 @@ function createCdpSender(ws: WebSocket) {
  return { send, closeWithError };
 }

-export async function fetchJson<T>(url: string, timeoutMs = 1500): Promise<T> {
+export async function fetchJson<T>(url: string, timeoutMs = 1500, init?: RequestInit): Promise<T> {
  const ctrl = new AbortController();
  const t = setTimeout(() => ctrl.abort(), timeoutMs);
  try {
-    const res = await fetch(url, { signal: ctrl.signal });
+    const headers = getHeadersWithAuth(url, (init?.headers as Record<string, string>) || {});
+    const res = await fetch(url, { ...init, headers, signal: ctrl.signal });
    if (!res.ok) throw new Error(`HTTP ${res.status}`);
    return (await res.json()) as T;
  } finally {
@@ -87,11 +111,28 @@ export async function fetchJson<T>(url: string, timeoutMs = 1500): Promise<T> {
  }
 }

+export async function fetchOk(url: string, timeoutMs = 1500, init?: RequestInit): Promise<void> {
+  const ctrl = new AbortController();
+  const t = setTimeout(() => ctrl.abort(), timeoutMs);
+  try {
+    const headers = getHeadersWithAuth(url, (init?.headers as Record<string, string>) || {});
+    const res = await fetch(url, { ...init, headers, signal: ctrl.signal });
+    if (!res.ok) throw new Error(`HTTP ${res.status}`);
+  } finally {
+    clearTimeout(t);
+  }
+}
+
 export async function withCdpSocket<T>(
  wsUrl: string,
  fn: (send: CdpSendFn) => Promise<T>,
+  opts?: { headers?: Record<string, string> },
 ): Promise<T> {
-  const ws = new WebSocket(wsUrl, { handshakeTimeout: 5000 });
+  const headers = getHeadersWithAuth(wsUrl, opts?.headers ?? {});
+  const ws = new WebSocket(wsUrl, {
+    handshakeTimeout: 5000,
+    ...(Object.keys(headers).length ? { headers } : {}),
+  });
  const { send, closeWithError } = createCdpSender(ws);

  const openPromise = new Promise<void>((resolve, reject) => {
@@ -165,4 +165,20 @@ describe("cdp", () => {
    );
    expect(normalized).toBe("ws://example.com:9222/devtools/browser/ABC");
  });
+
+  it("propagates auth and query params onto normalized websocket URLs", () => {
+    const normalized = normalizeCdpWsUrl(
+      "ws://127.0.0.1:9222/devtools/browser/ABC",
+      "https://user:pass@example.com?token=abc",
+    );
+    expect(normalized).toBe("wss://user:pass@example.com/devtools/browser/ABC?token=abc");
+  });
+
+  it("upgrades ws to wss when CDP uses https", () => {
+    const normalized = normalizeCdpWsUrl(
+      "ws://production-sfo.browserless.io",
+      "https://production-sfo.browserless.io?token=abc",
+    );
+    expect(normalized).toBe("wss://production-sfo.browserless.io/?token=abc");
+  });
 });
@@ -1,4 +1,6 @@
-import { fetchJson, isLoopbackHost, withCdpSocket } from "./cdp.helpers.js";
+import { appendCdpPath, fetchJson, isLoopbackHost, withCdpSocket } from "./cdp.helpers.js";
+
+export { appendCdpPath, fetchJson, fetchOk, getHeadersWithAuth } from "./cdp.helpers.js";

 export function normalizeCdpWsUrl(wsUrl: string, cdpUrl: string): string {
  const ws = new URL(wsUrl);
@@ -9,6 +11,16 @@ export function normalizeCdpWsUrl(wsUrl: string, cdpUrl: string): string {
    if (cdpPort) ws.port = cdpPort;
    ws.protocol = cdp.protocol === "https:" ? "wss:" : "ws:";
  }
+  if (cdp.protocol === "https:" && ws.protocol === "ws:") {
+    ws.protocol = "wss:";
+  }
+  if (!ws.username && !ws.password && (cdp.username || cdp.password)) {
+    ws.username = cdp.username;
+    ws.password = cdp.password;
+  }
+  for (const [key, value] of cdp.searchParams.entries()) {
+    if (!ws.searchParams.has(key)) ws.searchParams.append(key, value);
+  }
  return ws.toString();
 }

@@ -68,8 +80,10 @@ export async function createTargetViaCdp(opts: {
  cdpUrl: string;
  url: string;
 }): Promise<{ targetId: string }> {
-  const base = opts.cdpUrl.replace(/\/$/, "");
-  const version = await fetchJson<{ webSocketDebuggerUrl?: string }>(`${base}/json/version`, 1500);
+  const version = await fetchJson<{ webSocketDebuggerUrl?: string }>(
+    appendCdpPath(opts.cdpUrl, "/json/version"),
+    1500,
+  );
  const wsUrlRaw = String(version?.webSocketDebuggerUrl ?? "").trim();
  const wsUrl = wsUrlRaw ? normalizeCdpWsUrl(wsUrlRaw, opts.cdpUrl) : "";
  if (!wsUrl) throw new Error("CDP /json/version missing webSocketDebuggerUrl");
@@ -24,10 +24,12 @@ describe("browser default executable detection", () => {
  it("prefers default Chromium browser on macOS", async () => {
    vi.mocked(execFileSync).mockImplementation((cmd, args) => {
      const argsStr = Array.isArray(args) ? args.join(" ") : "";
-      if (cmd === "/usr/bin/osascript" && argsStr.includes("id of application")) {
-        return "com.google.Chrome";
+      if (cmd === "/usr/bin/plutil" && argsStr.includes("LSHandlers")) {
+        return JSON.stringify([
+          { LSHandlerURLScheme: "http", LSHandlerRoleAll: "com.google.Chrome" },
+        ]);
      }
-      if (cmd === "/usr/bin/osascript" && argsStr.includes("POSIX path")) {
+      if (cmd === "/usr/bin/osascript" && argsStr.includes("path to application id")) {
        return "/Applications/Google Chrome.app";
      }
      if (cmd === "/usr/bin/defaults") {
@@ -35,9 +37,11 @@ describe("browser default executable detection", () => {
      }
      return "";
    });
-    vi.mocked(fs.existsSync).mockImplementation((p) =>
-      String(p).includes("/Applications/Google Chrome.app/Contents/MacOS/Google Chrome"),
-    );
+    vi.mocked(fs.existsSync).mockImplementation((p) => {
+      const value = String(p);
+      if (value.includes("com.apple.launchservices.secure.plist")) return true;
+      return value.includes("/Applications/Google Chrome.app/Contents/MacOS/Google Chrome");
+    });

    const { resolveBrowserExecutableForPlatform } = await import("./chrome.executables.js");
    const exe = resolveBrowserExecutableForPlatform(
@@ -52,14 +56,18 @@ describe("browser default executable detection", () => {
  it("falls back when default browser is non-Chromium on macOS", async () => {
    vi.mocked(execFileSync).mockImplementation((cmd, args) => {
      const argsStr = Array.isArray(args) ? args.join(" ") : "";
-      if (cmd === "/usr/bin/osascript" && argsStr.includes("id of application")) {
-        return "com.apple.Safari";
+      if (cmd === "/usr/bin/plutil" && argsStr.includes("LSHandlers")) {
+        return JSON.stringify([
+          { LSHandlerURLScheme: "http", LSHandlerRoleAll: "com.apple.Safari" },
+        ]);
      }
      return "";
    });
-    vi.mocked(fs.existsSync).mockImplementation((p) =>
-      String(p).includes("Google Chrome.app/Contents/MacOS/Google Chrome"),
-    );
+    vi.mocked(fs.existsSync).mockImplementation((p) => {
+      const value = String(p);
+      if (value.includes("com.apple.launchservices.secure.plist")) return true;
+      return value.includes("Google Chrome.app/Contents/MacOS/Google Chrome");
+    });

    const { resolveBrowserExecutableForPlatform } = await import("./chrome.executables.js");
    const exe = resolveBrowserExecutableForPlatform(
@@ -95,12 +95,17 @@ function exists(filePath: string) {
  }
 }

-function execText(command: string, args: string[], timeoutMs = 1200): string | null {
+function execText(
+  command: string,
+  args: string[],
+  timeoutMs = 1200,
+  maxBuffer = 1024 * 1024,
+): string | null {
  try {
    const output = execFileSync(command, args, {
      timeout: timeoutMs,
      encoding: "utf8",
-      maxBuffer: 1024 * 1024,
+      maxBuffer,
    });
    return String(output ?? "").trim() || null;
  } catch {
@@ -114,7 +119,12 @@ function inferKindFromIdentifier(identifier: string): BrowserExecutable["kind"]
  if (id.includes("edge")) return "edge";
  if (id.includes("chromium")) return "chromium";
  if (id.includes("canary")) return "canary";
-  if (id.includes("opera") || id.includes("vivaldi") || id.includes("yandex") || id.includes("thebrowser")) {
+  if (
+    id.includes("opera") ||
+    id.includes("vivaldi") ||
+    id.includes("yandex") ||
+    id.includes("thebrowser")
+  ) {
    return "chromium";
  }
  return "chrome";
@@ -126,13 +136,12 @@ function inferKindFromExecutableName(name: string): BrowserExecutable["kind"] {
  if (lower.includes("edge") || lower.includes("msedge")) return "edge";
  if (lower.includes("chromium")) return "chromium";
  if (lower.includes("canary") || lower.includes("sxs")) return "canary";
-  if (lower.includes("opera") || lower.includes("vivaldi") || lower.includes("yandex")) return "chromium";
+  if (lower.includes("opera") || lower.includes("vivaldi") || lower.includes("yandex"))
+    return "chromium";
  return "chrome";
 }

-function detectDefaultChromiumExecutable(
-  platform: NodeJS.Platform,
-): BrowserExecutable | null {
+function detectDefaultChromiumExecutable(platform: NodeJS.Platform): BrowserExecutable | null {
  if (platform === "darwin") return detectDefaultChromiumExecutableMac();
  if (platform === "linux") return detectDefaultChromiumExecutableLinux();
  if (platform === "win32") return detectDefaultChromiumExecutableWindows();
@@ -140,14 +149,12 @@ function detectDefaultChromiumExecutable(
 }

 function detectDefaultChromiumExecutableMac(): BrowserExecutable | null {
-  const bundleId = execText("/usr/bin/osascript", [
-    "-e",
-    'id of application (path to default application for URL "http://example.com")',
-  ]);
-  if (!bundleId || !CHROMIUM_BUNDLE_IDS.has(bundleId.trim())) return null;
+  const bundleId = detectDefaultBrowserBundleIdMac();
+  if (!bundleId || !CHROMIUM_BUNDLE_IDS.has(bundleId)) return null;
+
  const appPathRaw = execText("/usr/bin/osascript", [
    "-e",
-    'POSIX path of (path to default application for URL "http://example.com")',
+    `POSIX path of (path to application id "${bundleId}")`,
  ]);
  if (!appPathRaw) return null;
  const appPath = appPathRaw.trim().replace(/\/$/, "");
@@ -162,6 +169,45 @@ function detectDefaultChromiumExecutableMac(): BrowserExecutable | null {
  return { kind: inferKindFromIdentifier(bundleId), path: exePath };
 }

+function detectDefaultBrowserBundleIdMac(): string | null {
+  const plistPath = path.join(
+    os.homedir(),
+    "Library/Preferences/com.apple.LaunchServices/com.apple.launchservices.secure.plist",
+  );
+  if (!exists(plistPath)) return null;
+  const handlersRaw = execText(
+    "/usr/bin/plutil",
+    ["-extract", "LSHandlers", "json", "-o", "-", "--", plistPath],
+    2000,
+    5 * 1024 * 1024,
+  );
+  if (!handlersRaw) return null;
+  let handlers: unknown;
+  try {
+    handlers = JSON.parse(handlersRaw);
+  } catch {
+    return null;
+  }
+  if (!Array.isArray(handlers)) return null;
+
+  const resolveScheme = (scheme: string) => {
+    let candidate: string | null = null;
+    for (const entry of handlers) {
+      if (!entry || typeof entry !== "object") continue;
+      const record = entry as Record<string, unknown>;
+      if (record.LSHandlerURLScheme !== scheme) continue;
+      const role =
+        (typeof record.LSHandlerRoleAll === "string" && record.LSHandlerRoleAll) ||
+        (typeof record.LSHandlerRoleViewer === "string" && record.LSHandlerRoleViewer) ||
+        null;
+      if (role) candidate = role;
+    }
+    return candidate;
+  };
+
+  return resolveScheme("http") ?? resolveScheme("https");
+}
+
 function detectDefaultChromiumExecutableLinux(): BrowserExecutable | null {
  const desktopId =
    execText("xdg-settings", ["get", "default-web-browser"]) ||
@@ -185,8 +231,7 @@ function detectDefaultChromiumExecutableLinux(): BrowserExecutable | null {
 function detectDefaultChromiumExecutableWindows(): BrowserExecutable | null {
  const progId = readWindowsProgId();
  const command =
-    (progId ? readWindowsCommandForProgId(progId) : null) ||
-    readWindowsCommandForProgId("http");
+    (progId ? readWindowsCommandForProgId(progId) : null) || readWindowsCommandForProgId("http");
  if (!command) return null;
  const expanded = expandWindowsEnvVars(command);
  const exePath = extractWindowsExecutablePath(expanded);
@@ -243,7 +288,7 @@ function splitExecLine(line: string): string[] {
  let quoteChar = "";
  for (let i = 0; i < line.length; i += 1) {
    const ch = line[i];
-    if ((ch === "\"" || ch === "'") && (!inQuotes || ch === quoteChar)) {
+    if ((ch === '"' || ch === "'") && (!inQuotes || ch === quoteChar)) {
      if (inQuotes) {
        inQuotes = false;
        quoteChar = "";
@@ -300,7 +345,7 @@ function readWindowsCommandForProgId(progId: string): string | null {
 function expandWindowsEnvVars(value: string): string {
  return value.replace(/%([^%]+)%/g, (_match, name) => {
    const key = String(name ?? "").trim();
-    return key ? process.env[key] ?? `%${key}%` : _match;
+    return key ? (process.env[key] ?? `%${key}%`) : _match;
  });
 }

@@ -7,7 +7,8 @@ import WebSocket from "ws";
 import { ensurePortAvailable } from "../infra/ports.js";
 import { createSubsystemLogger } from "../logging.js";
 import { CONFIG_DIR } from "../utils.js";
-import { normalizeCdpWsUrl } from "./cdp.js";
+import { getHeadersWithAuth, normalizeCdpWsUrl } from "./cdp.js";
+import { appendCdpPath } from "./cdp.helpers.js";
 import {
  type BrowserExecutable,
  resolveBrowserExecutableForPlatform,
@@ -71,9 +72,10 @@ async function fetchChromeVersion(cdpUrl: string, timeoutMs = 500): Promise<Chro
  const ctrl = new AbortController();
  const t = setTimeout(() => ctrl.abort(), timeoutMs);
  try {
-    const base = cdpUrl.replace(/\/$/, "");
-    const res = await fetch(`${base}/json/version`, {
+    const versionUrl = appendCdpPath(cdpUrl, "/json/version");
+    const res = await fetch(versionUrl, {
      signal: ctrl.signal,
+      headers: getHeadersWithAuth(versionUrl),
    });
    if (!res.ok) return null;
    const data = (await res.json()) as ChromeVersion;
@@ -98,7 +100,11 @@ export async function getChromeWebSocketUrl(

 async function canOpenWebSocket(wsUrl: string, timeoutMs = 800): Promise<boolean> {
  return await new Promise<boolean>((resolve) => {
-    const ws = new WebSocket(wsUrl, { handshakeTimeout: timeoutMs });
+    const headers = getHeadersWithAuth(wsUrl);
+    const ws = new WebSocket(wsUrl, {
+      handshakeTimeout: timeoutMs,
+      ...(Object.keys(headers).length ? { headers } : {}),
+    });
    const timer = setTimeout(
      () => {
        try {
Author	SHA1	Message	Date
Peter Steinberger	01aa7eba21	chore: ignore local identity files (#1001 ) (thanks @gerardward2007)	2026-01-16 09:58:21 +00:00
gerardward2007andClaude Opus 4.5	9479fded34	chore: ignore local identity files (IDENTITY.md, USER.md) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>	2026-01-16 09:54:42 +00:00
Peter Steinberger	b969c216fc	docs: refresh unreleased highlights	2026-01-16 09:51:27 +00:00
Peter Steinberger	f51a4d2aca	docs: sort unreleased changelog	2026-01-16 09:46:12 +00:00
Peter Steinberger	30b3a9de30	fix: drop oauth status from session status	2026-01-16 09:39:12 +00:00
Peter Steinberger	0391f6553b	fix: correct minimax usage + show reset	2026-01-16 09:36:45 +00:00
Peter Steinberger	d0c986c4f0	feat: warn on weak model tiers	2026-01-16 09:34:37 +00:00
Peter Steinberger	384028e12e	refactor: unify session reset helper	2026-01-16 09:33:39 +00:00
Peter Steinberger	4c14d6c8db	chore: format web monitor inbox tests	2026-01-16 09:26:18 +00:00
Peter Steinberger	6fa437613b	fix: delete transcripts on role ordering reset	2026-01-16 09:20:16 +00:00
Peter Steinberger	949fa1051f	fix: wire markdown variant renderer	2026-01-16 09:19:25 +00:00
Peter Steinberger	4965727f39	chore: run format and fix sandbox browser timeouts	2026-01-16 09:18:58 +00:00
Peter Steinberger	7c34883267	refactor: consolidate chat markdown rendering	2026-01-16 09:16:43 +00:00
Peter Steinberger	072c3dc55c	fix: suppress WhatsApp pairing replies for historical DMs	2026-01-16 09:10:44 +00:00
Peter Steinberger	83b3875131	docs: add control UI unauthorized FAQ hint	2026-01-16 09:07:20 +00:00
Peter Steinberger	a35083808c	chore: update macOS package lock	2026-01-16 09:06:23 +00:00
Peter Steinberger	3d3ec9d972	docs: add browserless remote CDP example	2026-01-16 09:05:30 +00:00
Peter Steinberger	9838a2850f	fix: reset sessions after role ordering conflicts	2026-01-16 09:04:04 +00:00
Peter Steinberger	6c6bc6ff1c	fix: add control UI auth guidance	2026-01-16 09:03:02 +00:00
Peter Steinberger	1791c1a765	feat: render native chat markdown via Textual	2026-01-16 09:02:27 +00:00
Peter Steinberger	6e53c061ff	fix: tune remote CDP timeouts	2026-01-16 09:01:25 +00:00
Peter Steinberger	1773f8aea2	fix: upgrade ws to wss for https CDP	2026-01-16 08:44:08 +00:00
Peter Steinberger	52bdf57743	Merge pull request #880 from mkbehr/openai-image-gen-fix fix: Make openai-image-gen skill use appropriate defaults per model.	2026-01-16 08:42:13 +00:00
Peter Steinberger	b3ab24eb8e	fix: align image-gen defaults (#880 ) (thanks @mkbehr)	2026-01-16 08:41:23 +00:00
Michael BehrandClaude Opus 4.5	6ac1c1d6ea	fix(openai-image-gen): use correct file extension for output format When --output-format is specified for GPT models, save files with the correct extension (.webp, .jpeg, or .png) instead of always using .png. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>	2026-01-16 08:37:57 +00:00
Michael BehrandClaude Opus 4.5	7655a501d0	feat(openai-image-gen): add model-specific parameter support - Auto-detect model and apply appropriate defaults for size/quality - Add --background, --output-format, and --style parameters - Enforce dall-e-3 count=1 limitation with automatic adjustment - Omit quality parameter for dall-e-2 (not supported) - Document model-specific parameters and supported values 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>	2026-01-16 08:37:57 +00:00
Peter Steinberger	3b1b14b0b1	Merge pull request #895 from mukhtharcm/feat/chrome-browser-improvements feat(browser): add support for authenticated remote browser profiles	2026-01-16 08:32:02 +00:00
Peter Steinberger	bf15c87d2b	fix: support authenticated remote CDP URLs (#895 ) (thanks @mukhtharcm)	2026-01-16 08:31:51 +00:00
Peter Steinberger	2dae4d382f	Merge pull request #860 from nachoiacovino/feat/telegram-custom-commands feat(telegram): support custom commands in config	2026-01-16 08:26:26 +00:00
Peter Steinberger	e9a47a02d1	fix: stabilize macOS audio test and default browser detection	2026-01-16 08:25:39 +00:00
Peter SteinbergerandNacho Iacovino	929666a8c8	fix: add telegram custom commands (#860 ) (thanks @nachoiacovino) Co-authored-by: Nacho Iacovino <50103937+nachoiacovino@users.noreply.github.com>	2026-01-16 08:22:09 +00:00
Muhammed Mukhthar CMandPeter Steinberger	cd409e5667	fix: exclude google-antigravity from history downgrade hack (#894 ) * Agent: exclude google-antigravity from history downgrade hack * Lint: fix formatting in test * Lint: formatting and unused vars in test * fix: preserve google-antigravity tool calls (#894) (thanks @mukhtharcm) --------- Co-authored-by: Peter Steinberger <steipete@gmail.com>	2026-01-16 08:14:56 +00:00
Muhammed Mukhthar CM	8e80823b03	test(browser): fix missing getHeadersWithAuth mock in server tests	2026-01-16 08:11:00 +00:00
Muhammed Mukhthar CM	319afd192d	feat(browser): add support for remote playwright websocket auth	2026-01-16 08:11:00 +00:00
Muhammed Mukhthar CM	6e0daf0936	feat(browser): add support for authenticated remote CDP profiles	2026-01-16 08:10:32 +00:00
Peter Steinberger	d0cb4e092f	Merge pull request #845 from MatthieuBizien/fix/issue-841-openrouter-gemini Agents: sanitize OpenRouter Gemini thoughtSignature	2026-01-16 08:03:40 +00:00
Peter Steinberger	f5a881c99d	fix: port OpenRouter Gemini sanitization to split files (#845 ) (thanks @MatthieuBizien)	2026-01-16 08:02:56 +00:00
Peter Steinberger	66377fc030	fix: update macOS IPC tests	2026-01-16 07:58:35 +00:00
Matthieu Bizien	d8d295b0b3	Changelog: thank PR #845	2026-01-16 07:51:49 +00:00
Matthieu Bizien	ef36e24522	Agents: sanitize OpenRouter Gemini thoughtSignature	2026-01-16 07:51:49 +00:00
Palash OswalandPeter Steinberger	d43d4fcced	Gateway auth: accept local Tailscale Serve hostnames and tailnet IPs (#885 ) * Gateway auth: accept local Tailscale Serve hostnames and tailnet IPs * fix: allow local Tailscale Serve hostnames (#885) (thanks @oswalpalash) --------- Co-authored-by: Peter Steinberger <steipete@gmail.com>	2026-01-16 07:51:25 +00:00
Peter Steinberger	d42b69df74	Merge pull request #982 from wes-davis/fix/gateway-connection-diagnostics macOS: keep gateway connected (stop port flapping)	2026-01-16 07:36:46 +00:00
Peter Steinberger	1ec1f6dcbf	fix: sync remote ssh targets	2026-01-16 07:33:15 +00:00
Peter Steinberger	e96b939732	feat: add system.which bin probe	2026-01-16 07:31:41 +00:00
Peter Steinberger	e479c870fd	fix: handle MiniMax coding plan usage payloads	2026-01-16 07:28:48 +00:00
Peter Steinberger	f2db894685	Merge pull request #992 from tyler6204/fix/tool-typing-race-condition fix: send text between tool calls to channel immediately	2026-01-16 07:26:10 +00:00
Peter Steinberger	4f03283126	docs: add npm 1password publish steps	2026-01-16 07:08:07 +00:00
Peter Steinberger	ed7dec0975	feat: use dropdowns for access controls	2026-01-16 07:00:05 +00:00
Peter Steinberger	fbb3da506f	docs: clarify onboarding + sessions + heartbeats	2026-01-16 06:57:54 +00:00
Peter Steinberger	dfa6c5c2b3	fix(google): scrub tool schemas for gemini	2026-01-16 06:57:54 +00:00
Peter Steinberger	028eed5fe8	fix(browser): surface detection details and docs	2026-01-16 06:57:54 +00:00
Peter Steinberger	2b16a87f04	feat: add config get/set/unset helpers	2026-01-16 06:57:54 +00:00
Peter Steinberger	731049936d	Merge pull request #870 from JDIVE/fix/discord-actions-schema fix(config): allow discord action flags in schema	2026-01-16 06:50:50 +00:00
Peter Steinberger	5a5b058ba0	fix: update discord action defaults (#870 ) (thanks @JDIVE)	2026-01-16 06:50:27 +00:00
Jamie OpenshawandClaude	72f28be648	fix(config): allow discord action flags in schema Ensure discord action flags survive config validation. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>	2026-01-16 06:48:25 +00:00
Peter Steinberger	ff645524d8	docs: move troubleshooting items from faq	2026-01-16 06:35:13 +00:00
Peter Steinberger	c534390bc0	chore: update a2ui bundle	2026-01-16 06:35:05 +00:00
Peter Steinberger	c5003e5441	fix: clear lint blockers	2026-01-16 06:35:05 +00:00
Peter Steinberger	0b3ebb0c63	test: stabilize slow and flaky tests	2026-01-16 06:24:58 +00:00
Peter Steinberger	23981496f9	fix: resolve bridge warnings	2026-01-16 06:15:45 +00:00
Peter Steinberger	f2e425dc2b	Merge pull request #995 from roshanasingh4/fix/systemd-execstart-whitespace Fix systemd ExecStart whitespace parsing	2026-01-16 06:03:35 +00:00
Peter Steinberger	e48d68bbc7	Merge pull request #993 from cpojer/reminder-improvement Improve reminder text generation.	2026-01-16 06:03:04 +00:00
Peter Steinberger	842fc8d08b	fix: repair launchd status parsing	2026-01-16 06:01:28 +00:00
Peter Steinberger	54ec14262b	feat: add plugin update tracking	2026-01-16 05:55:05 +00:00
Peter Steinberger	d0c70178e0	docs: fix FAQ reset anchor	2026-01-16 05:54:56 +00:00
Peter Steinberger	3bc9c330eb	fix: unblock mac node bridge TLS	2026-01-16 05:50:55 +00:00
Peter Steinberger	b7fcc8584f	style: apply swiftformat fixes	2026-01-16 05:42:18 +00:00
Roshan Singh	fa9aafce83	Fix systemd ExecStart parsing whitespace	2026-01-16 05:25:13 +00:00
Tyler Yust	0d5dec4c66	fix: handle async tool start handler rejections Add .catch() to handleToolExecutionStart call to prevent unhandled promise rejections when onAgentEvent or typing signaling fails.	2026-01-15 21:10:52 -08:00
cpojer	b2d5889f6e	Improve reminder text generation.	2026-01-16 14:03:17 +09:00
Tyler Yust	2ee71e4154	fix: send text between tool calls to channel immediately Previously, when block streaming was disabled (the default), text generated between tool calls would only appear after all tools completed. This was because onBlockReply wasn't passed to the subscription when block streaming was off, so flushBlockReplyBuffer() before tool execution did nothing. Now onBlockReply is always passed, and when block streaming is disabled, block replies are sent directly during tool flush. Directly sent payloads are tracked to avoid duplicates in final payloads. Also fixes a race condition where tool summaries could be emitted before the typing indicator started by awaiting onAgentEvent in tool handlers.	2026-01-15 20:55:52 -08:00
Wes	509215e935	macOS: stop flapping gateway port	2026-01-15 17:12:14 -08:00
Wes	f726656d1e	macOS: start gateway before app in restart-mac	2026-01-15 17:12:14 -08:00
Wes	30d3e1da21	CLI: fix status --all gateway auth selection	2026-01-15 17:12:14 -08:00