* test(cli): use lightweight clears in daemon lifecycle setup

* test(models): use lightweight clears in shared config setup

* test(agents): use lightweight clears for stable subagent announce defaults

* Sessions: persist prompt-token totals without usage

* fix(security): normalize hook auth rate-limit client keys

* refactor(cli): dedupe skills command report loading

* refactor(cli): dedupe channel auth resolution flow

* refactor(cli): dedupe allowlist command wiring

* test(cli): dedupe update restart fallback scenario setup

* test(cli): dedupe cron shared test fixtures

* refactor(cli): extract fish completion line builders

* test(cli): share nodes ios fixture helpers

* refactor(cli): share npm install metadata helpers

* refactor(cli): share pinned npm install record helper

* refactor(slack): dedupe modal lifecycle interaction handlers

* refactor(commands): share preview streaming migration logic

* test(gateway): reuse last agent command assertion helper

* test(discord): share provider lifecycle test harness

* test(discord): share thread binding sweep fixtures

* test(infra): dedupe shell env fallback test setup

* refactor(discord): dedupe voice command runtime checks

* test(discord): share model picker fallback fixtures

* test(discord): share message handler draft fixtures

* test(discord): share resolve-users guild probe fixture

* test(inbound): share dispatch capture mock across channels

* test(security): dedupe external marker sanitization assertions

* test(wizard): share onboarding prompter scaffold

* test(memory): share memory-tool manager mock fixture

* test(subagents): dedupe focus thread setup fixtures

* test(auth-profiles): dedupe cleared-state assertions

* test(memory): share short-timeout test helper

* test(outbound): share resolveOutboundTarget test suite

* test(auth-profiles): dedupe oauth mode resolution setup

* test(gateway): dedupe transcript seed fixtures in fs session tests

* refactor(text): share code-region parsing for reasoning tags

* refactor(node-host): share invoke type definitions

* refactor(logging): share node createRequire resolution

* test(models): dedupe auth-sync command assertions

* test(pi): share overflow-compaction test setup

* test(discord): dedupe guild permission route mocks

* refactor(config): dedupe legacy stream-mode migration paths

* test(gateway): dedupe tailscale header auth fixtures

* test(browser): dedupe relay probe server scaffolding

* test(cron): dedupe delivered-status run scaffolding

* test(gateway): dedupe control-ui not-found fixture assertions

* test(gateway): dedupe openai context assertions

* test(config): dedupe traversal include assertions

* test(config): dedupe nested redaction round-trip assertions

* test(gateway): reuse shared openai timeout e2e helpers

* test(gateway): dedupe chat history transcript helpers

* test(gateway): dedupe canvas ws connect assertions

* test(hooks): dedupe unsupported npm spec assertion

* test(agent): reuse isolated agent mock setup

* test(utils): share temp-dir helper across cli and web tests

* test(browser): dedupe generated-token persistence assertions

* test(browser): dedupe pw-session playwright mock wiring

* test(agents): dedupe spawn-hook wait mocks and add readiness error coverage

* test(agents): dedupe sanitize-session-history copilot fixtures

* test: dedupe lifecycle oauth and prompt-limit fixtures

* refactor(agents): share volc model catalog helpers

* refactor(agents): reuse shared tool-policy base helpers

* refactor: eliminate remaining duplicate blocks across draft streams and tests

* refactor(core): dedupe gateway runtime and config tests

* refactor(channels): dedupe message routing and telegram helpers

* refactor(agents): dedupe plugin hooks and test helpers

* chore: remove dead plugin hook loader

* fix(security): harden gateway command/audit guardrails

* test: dedupe telegram draft stream setup and extend state-dir env coverage

* Agents: drop stale pre-compaction usage snapshots

* docs(changelog): note next npm release for hook auth fix

* test(telegram): dedupe native-command test setup

* fix(gateway): block avatar symlink escapes

* test: dedupe cron and slack monitor test harness setup

* refactor(security): unify hook rate-limit and hook module loading

* test(gateway): dedupe loopback cases and trim setup resets

* test(agents): use lightweight clears in supervisor and session-status setup

* test(auto-reply): centralize subagent command test reset setup

* test(agents): centralize sessions tool gateway mock reset

* test(telegram): centralize native command session-meta mock setup

* test(browser): use lightweight clears in server lifecycle setup

* test(gateway): use lightweight clears in cron service setup

* test(commands): use lightweight clears in doctor memory search setup

* test(outbound): dedupe shared setup hooks in message e2e

* test(gateway): use lightweight clears in push handler setup

* test(gateway): use lightweight clears in node invoke wake setup

* test(gateway): use lightweight clears in node event setup

* test(gateway): use lightweight clears for hook cron run fences

* test(auto-reply): use lightweight clears in dispatch setup

* test(agents): use lightweight clears in sandbox browser create setup

* test(auto-reply): use lightweight clears in agent runner setup

* test(plugins): use lightweight clears in wired hooks setup

* test(gateway): use lightweight clears in client close setup

* test(ui): use lightweight clears in theme and telegram media retry setup

* test(agents): use lightweight clears in skills install e2e setup

* test(gateway): use lightweight clears for chat-b reply spy fences

* test(gateway): use lightweight clears for openai http agent fences

* test(gateway): use lightweight clears for openresponses agent fences

* test(core): use lightweight clears in update, child adapter, and copilot token setup

* test(agents): dedupe sessions_spawn e2e reset setup

* test(core): use lightweight clears in stable mock setup

* test(agents): dedupe sessions_spawn allowlist reset setup

* test(agents): drop redundant subagent registry cleanups

* test(core): trim redundant mock resets in heartbeat suites

* test(daemon): use lightweight clears in systemd mocks

* test(infra): use lightweight clears in update startup mocks

* test(gateway): use lightweight clears in agent handler tests

* test(infra): use lightweight clears in message action threading setup

* test(telegram): use lightweight clears in media handler setup

* test(commands): use lightweight clears in agents/channels setup

* fix: align draft/outbound typings and tests

* test: stabilize pw-session cdp mocking in parallel runs

* chore(docs): normalize security finding table formatting

* fix(ci): add explicit mock types in pw-session mock setup

* test(core): use lightweight clears in command and dispatch setup

* test(agents): use lightweight clears in skills/sandbox setup

* test(core): use lightweight clears in subagent and browser setup

* test(core): use lightweight clears in runtime and telegram setup

* test(core): trim redundant test resets and use mockClear

* test(slack): use lightweight clear in interactions modal-close case

* test(slack): avoid redundant reset in slash metadata wait case

* test(reply): replace heavy resets in media and runner helper specs

* test(agents): reduce reset overhead in session visibility and hooks specs

* test(subagents): lighten session delete mock reset in announce spec

* test(memory): prefer clear over reset in qmd spawn setup

* test(agents): keep targeted resets minimal in overflow retry spec

* chore: remove verified dead code paths

* test(core): reduce mock reset overhead across unit and e2e specs

* Agents: add fallback reply for tool-only completions

* test(core): trim reset usage in gateway and install source specs

* test(commands): use lightweight clears in config snapshot specs

* refactor(gateway)!: remove legacy v1 device-auth handshake

* test(subagents): use lightweight clears in sessions spawn suites

* test(core): continue mock reset reductions in auth, gateway, npm install

* test(core): continue reset-to-clear cleanup in subagent focus and web fetch

* test(config): use lightweight clear in session pruning e2e setup

* test(core): reduce reset overhead in messaging and agent e2e mocks

* test(core): tighten reset usage in auth, registry restart, and memory search

* fix: decouple owner display secret from gateway auth token

* chore: remove dead macos relay and daemon code

* test(core): use lightweight clear in cron, claude runner, and telegram delivery specs

* Agents/Subagents: honor subagent alsoAllow grants

* test(core): reduce mock reset overhead in targeted suites

* fix(security): block HOME and ZDOTDIR env override injection

* test(core): dedupe auth rotation and credential injection specs

* test(agents): dedupe subagent announce direct-send variants

* docs(changelog): add shell startup env override fix note

* chore(test): make shell-env trusted-shell assertion platform-aware

* test(commands): dedupe subagent status assertions

* fix: harden exec allowlist wrapper resolution

* test(agents): avoid full mock resets in cli credential specs

* chore(test): harden models status mock restoration

* test(core): dedupe command gating and trim announce reset overhead

* test(agents): unify hook thread-target announce assertions

* test(agents): collapse repeated announce direct-send scenarios

* test(reply): merge duplicate runReplyAgent streaming and fallback cases

* test(agents): use lightweight clear for active-run announce mock

* test(agents): remove overflow compaction mock reset dependency

* test(reply): use lightweight clears for runner-level mocks

* test(agents): consolidate repeated announce deferral and fallback matrices

* test(commands): replace subagent gateway reset with lightweight clear

* TUI: preserve RTL text order in terminal output

* docs(security): clarify dangerous control-ui bypass policy

* feat(security): warn on dangerous config flags at startup

* perf(test): bypass queue debounce in fast mode and tighten announce defaults

* fix(security): harden channel token and id generation

* refactor(security): unify secure id paths and guard weak patterns

* fix(gateway): remove hello-ok host and commit fields

* fix(security): block hook transform symlink escapes

* refactor: unify exec wrapper resolution and parity fixtures

* TUI: make Ctrl+C exit behavior reliably responsive

* test(heartbeat): dedupe sandbox/session helpers and collapse ack cases

* test(agents): simplify subagent announce suite imports and call assertions

* test(heartbeat): reuse shared temp sandbox in model override suite

* test(heartbeat): reuse shared sandbox for ghost reminder scenarios

* perf(test): compact heartbeat session fixture writes

* perf(test): shrink subagent announce fast-mode settle waits

* fix: use SID-based ACL classification for non-English Windows

* fix: detect zombie processes in isPidAlive on Linux

kill(pid, 0) succeeds for zombie processes, causing the gateway lock
to treat a zombie lock owner as alive. Read /proc/<pid>/status on
Linux to check for 'Z' (zombie) state before reporting the process
as alive. This prevents the lock from being held indefinitely by a
zombie process during gateway restart.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: release gateway lock before process.exit in run-loop

process.exit() called from inside an async IIFE bypasses the outer
try/finally block that releases the gateway lock. This leaves a stale
lock file pointing to a zombie PID, preventing the spawned child or
systemctl restart from acquiring the lock. Release the lock explicitly
before calling exit in both the restart-spawned and stop code paths.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: release gateway lock before spawning restart child

Move lock.release() before restartGatewayProcessWithFreshPid() so the
spawned child can immediately acquire the lock without racing against
a zombie parent. This eliminates the root cause of the restart loop
where the child times out waiting for a lock held by its now-dead parent.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: guard entry.ts top-level code with isMainModule to prevent duplicate gateway start

The bundler exports shared symbols from dist/entry.js, so other chunks
import it as a dependency. When dist/index.js is the actual entry point
(e.g. systemd service), lazy module loading eventually imports entry.js,
triggering its unguarded top-level code which calls runCli(process.argv)
a second time. This starts a duplicate gateway that fails on lock/port
contention and crashes the process with exit(1), causing a restart loop.

Wrap all top-level executable code in an isMainModule() check so it only
runs when entry.ts is the actual main module, not when imported as a
shared dependency by the bundler.

* fix: tighten gateway restart loop handling (openclaw#23416) (thanks @jeffwnli)

* chore: fix temp-path guard skip for *.test-helpers.ts

* fix: include modelByChannel in config validator allowedChannels

The hand-written config validator rejects `channels.modelByChannel` as
"unknown channel id: modelByChannel" even though the Zod schema, TypeScript
types, runtime code, and CLI docs all treat it as valid. The `defaults`
meta-key was already whitelisted but `modelByChannel` was missed when
the feature was added in 2026.2.21.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* also skip modelByChannel in plugin-auto-enable channel iteration

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: cover channels.modelByChannel validation/auto-enable

* fix: finalize modelByChannel validator landing (openclaw#23412) (thanks @ProspectOre)

* refactor: simplify windows ACL parsing and expand coverage

* refactor(gateway): simplify restart flow and expand lock tests

* refactor(plugin-sdk): unify channel dedupe primitives

* fix(acp): wait for gateway connection before processing ACP messages

- Move gateway.start() before AgentSideConnection creation
- Wait for hello message to confirm connection is established
- This fixes issues where messages were processed before gateway was ready

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: harden ACP gateway startup sequencing (openclaw#23390) (thanks @janckerchen)

* Memory/QMD: normalize Han-script BM25 search queries

* fix(stability): patch regex retries and timeout abort handling

* fix: handle intentional signal daemon shutdown on abort (openclaw#23379) (thanks @frankekn)

* refactor(signal): extract daemon lifecycle and typed exit handling

* Exec: fail closed when sandbox host is unavailable

* fix: harden exec sandbox fallback semantics (openclaw#23398) (thanks @bmendonca3)

* test: stabilize temp-path guard across runtimes (openclaw#23398)

* test: harden temp path guard detection (openclaw#23398)

* fix(feishu): avoid template tmpdir join in dedup state path (openclaw#23398)

* feat(feishu): persistent message deduplication to prevent duplicate replies

Closes openclaw#23369

Feishu may redeliver the same message during WebSocket reconnects or process
restarts.  The existing in-memory dedup map is lost on restart, so duplicates
slip through.

This adds a dual-layer dedup strategy:
- Memory cache (fast synchronous path, unchanged capacity)
- Filesystem store (~/.openclaw/feishu/dedup/) that survives restarts

TTL is extended from 30 min to 24 h.  Disk writes use atomic rename and
probabilistic cleanup to keep each per-account file under 10 k entries.
Disk errors are caught and logged — message handling falls back to
memory-only behaviour so it is never blocked.

* fix(feishu): address dedup race condition, namespace isolation, and cache staleness

- Prefix memoryCache keys with namespace to prevent cross-account false
  positives when different accounts receive the same message_id
- Add inflight tracking map to prevent TOCTOU race where concurrent
  async calls for the same message both pass the check and both proceed
- Remove expired-entry deletion from has() to avoid silent cache/disk
  divergence; actual cleanup happens probabilistically inside record()
- Add time-based cache invalidation (30s) to DedupStore.load() so
  external writes are eventually picked up
- Refresh cacheLoadedAt after flush() so we don't immediately re-read
  data we just wrote

Co-authored-by: Cursor <cursoragent@cursor.com>

* fix: tighten feishu dedupe boundary (openclaw#23377) (thanks @SidQin-cyber)

* Feat/logger support log level validation0222 (openclaw#23436)

* 1、环境变量**：新增 `OPENCLAW_LOG_LEVEL`，可取值 `silent|fatal|error|warn|info|debug|trace`。设置后同时覆盖**文件日志**与**控制台**的级别，优先级高于配置文件。
2、启动参数**：在 `openclaw gateway run` 上新增 `--log-level <level>`，对该次进程同时生效于文件与控制台；未传时仍使用环境变量或配置文件。

* fix(logging): make log-level override global and precedence-safe

---------

Co-authored-by: Peter Steinberger <steipete@gmail.com>

* fix(telegram): prevent update offset skipping queued updates (openclaw#23284)

Merged via /review-pr -> /prepare-pr -> /merge-pr.

Prepared head SHA: 92efaf9
Co-authored-by: frankekn <4488090+frankekn@users.noreply.github.com>
Co-authored-by: obviyus <22031114+obviyus@users.noreply.github.com>
Reviewed-by: @obviyus

* fix: stop hardcoded channel fallback and auto-pick sole configured channel (openclaw#23357) (thanks @lbo728)

Co-authored-by: lbo728 <extreme0728@gmail.com>

* docs(security): clarify workspace memory trust boundary

* Security: expand audit checks for mDNS and real-IP fallback

* fix: land security audit severity + temp-path guard fixes (openclaw#23428) (thanks @bmendonca3)

* test(heartbeat): use shared sandbox in sender target suite

* perf(test): compact remaining heartbeat fixture writes

* test(reply): align native trigger suite with fast-test fixture patterns

* perf(test): speed subagent announce retry polling in fast mode

* test(agents): dedupe auth profile rotation fixture setup

* perf(test): trim background abort settle waits and dedupe cmd fixture

* perf(test): trim nested subagent output wait floor in fast mode

* perf(test): lower fast-mode nested output wait floor to 80ms

* test(agents): remove dead shell-timeout override in safeBins suite

* perf(test): lower fast-mode nested output wait floor to 70ms

* perf(test): remove flaky transport timeout and dedupe safeBins checks

* perf(test): mock compact module in auth rotation e2e

* perf(test): reduce subagent announce fast-mode polling waits

* perf(test): lower subagent fast-mode wait floors

* perf(test): trim bash e2e sleep and poll windows

* perf(test): narrow pi-embedded runner e2e import path

* test: reclassify mocked runner/safe-bins suites as unit tests

* test: reclassify auth-profile-rotation suite as unit test

* test: reclassify mocked announce and sandbox suites as unit tests

* perf(test): tighten background abort timing windows

* test: reclassify sandbox merge and exec path suites as unit tests

* perf(test): speed up sessions_spawn lifecycle suite setup

* test: reclassify sessions_spawn lifecycle suite as unit test

* perf(test): reduce bash e2e wait windows

* fix(gateway): strip directive tags from non-streaming webchat broadcasts

Closes openclaw#23053

The streaming path already strips [[reply_to_current]] and other
directive tags via stripInlineDirectiveTagsForDisplay, but the
non-streaming broadcastChatFinal path and the chat.inject path
sent raw message content to webchat clients, causing tags to
appear in rendered messages after streaming completes.

* fix: add non-streaming directive-tag regression tests (openclaw#23298) (thanks @SidQin-cyber)

* test: reclassify skills suites from e2e to unit lane

* test: reclassify models-config suites from e2e to unit lane

* test: harden models-config env isolation list

* refactor: clarify strict loopback proxy audit rules

* fix(session): resolve agent session path with configured sessions dir

Co-authored-by: David Rudduck <david@rudduck.org.au>

* fix(telegram): classify undici fetch errors as recoverable for retry (openclaw#16699)

Merged via /review-pr -> /prepare-pr -> /merge-pr.

Prepared head SHA: 67b5bce
Co-authored-by: Glucksberg <80581902+Glucksberg@users.noreply.github.com>
Co-authored-by: obviyus <22031114+obviyus@users.noreply.github.com>
Reviewed-by: @obviyus

* fix(config): add missing comment field to BindingsSchema

Strict validation (added in d1e9490) rejects the legitimate 'comment'
field on bindings. This field is used for annotations in config files.

Changes:
- BindingsSchema: added comment: z.string().optional()
- AgentBinding type: added comment?: string

Fixes openclaw#23385

* fix: add bindings comment regression test (openclaw#23458) (thanks @echoVic)

* fix(bluebubbles): treat null privateApiStatus as disabled, not enabled

Bug: privateApiStatus cache expires after 10 minutes, returning null.
The check '!== false' treats null as truthy, causing 500 errors when
trying to use Private API features that aren't actually available.

Root cause: In JavaScript, null !== false evaluates to true.

Fix: Changed all checks from '!== false' to '=== true', so null (cache
expired/unknown) is treated as disabled (safe default).

Files changed:
- extensions/bluebubbles/src/send.ts (line 376)
- extensions/bluebubbles/src/monitor-processing.ts (line 423)
- extensions/bluebubbles/src/attachments.ts (lines 210, 220)

Fixes openclaw#23393

* fix: align BlueBubbles private-api null fallback + warning (openclaw#23459) (thanks @echoVic)

* refactor(session): centralize transcript path option resolution

* fix: add operator.read and operator.write to default CLI scopes (openclaw#22582)

Merged via /review-pr -> /prepare-pr -> /merge-pr.

Prepared head SHA: 8569fc8
Co-authored-by: YuzuruS <1485195+YuzuruS@users.noreply.github.com>
Co-authored-by: obviyus <22031114+obviyus@users.noreply.github.com>
Reviewed-by: @obviyus

* feat(workspace): add PROFILE-<name>.md bootstrap file support

When OPENCLAW_PROFILE is set (and not "default"), automatically load a
PROFILE-<profileName>.md file from the workspace as an additional
bootstrap context file. This gives each profile instance its own
personality/context overlay without needing hook configuration.

Changes:
- Add isProfileBootstrapName() helper to validate PROFILE-*.md pattern
- Update loadWorkspaceBootstrapFiles() to load profile file when env var is set
- Insert profile file in correct order (after USER.md, before HEARTBEAT.md)
- Update loadExtraBootstrapFiles() to accept PROFILE-*.md filenames
- Update filterBootstrapFilesForSession() to preserve profile files in subagent/cron sessions
- Widen WorkspaceBootstrapFileName type to include dynamic profile filenames
- Add comprehensive test coverage for all profile file scenarios
- Update bootstrap-extra-files hook documentation

The profile file is optional - if it doesn't exist, it's silently skipped
without adding a [MISSING] marker. This makes it zero-config for
multi-instance setups like hive clusters.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* ci: add promoted release workflow for v*-turq.* tags

---------

Co-authored-by: Peter Steinberger <steipete@gmail.com>
Co-authored-by: Vignesh Natarajan <vigneshnatarajan92@gmail.com>
Co-authored-by: SK Akram <skcodewizard786@gmail.com>
Co-authored-by: jeffr <jeffr@local>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: pickaxe <54486432+ProspectOre@users.noreply.github.com>
Co-authored-by: janckerchen <janckerchen@gmail.com>
Co-authored-by: Frank Yang <frank.ekn@gmail.com>
Co-authored-by: Brian Mendonca <brianmendonca@Brians-MacBook-Air.local>
Co-authored-by: SidQin-cyber <sidqin0410@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: maweibin <532282155@qq.com>
Co-authored-by: frankekn <4488090+frankekn@users.noreply.github.com>
Co-authored-by: obviyus <22031114+obviyus@users.noreply.github.com>
Co-authored-by: lbo728 <extreme0728@gmail.com>
Co-authored-by: David Rudduck <david@rudduck.org.au>
Co-authored-by: Glucksberg <80581902+Glucksberg@users.noreply.github.com>
Co-authored-by: echoVic <echoVic@users.noreply.github.com>
Co-authored-by: Yuzuru Suzuki <navitima@gmail.com>
Co-authored-by: YuzuruS <1485195+YuzuruS@users.noreply.github.com>

…jeffwnli)

…jeffwnli)

…jeffwnli)

…jeffwnli)

(cherry picked from commit dd07c06)

# Conflicts:
#	CHANGELOG.md
#	src/cli/gateway-cli/run-loop.test.ts
#	src/cli/gateway-cli/run-loop.ts
#	src/infra/infra-parsing.test.ts

…jeffwnli)

(cherry picked from commit dd07c06)

# Conflicts:
#	CHANGELOG.md
#	src/cli/gateway-cli/run-loop.test.ts
#	src/cli/gateway-cli/run-loop.ts
#	src/infra/infra-parsing.test.ts

…jeffwnli)