…xt (openclaw#91478)

Move Telegram inbound authorization into a single pre-cache gate so unauthorized DM text is never recorded into the reply-chain cache or dispatch-dedupe state before allowlist/pairing checks run. The gate covers fresh messages, edited messages, group edits, and edited channel posts; edits are authorized silently and never trigger pairing challenges, and requireTopic root DMs are dropped before pairing challenges.

Fixes openclaw#91209.

Thanks @sallyom!

….chunk

…port

…plicit config (openclaw#91361)

Merged via squash.

Prepared head SHA: ac54596
Co-authored-by: wangmiao0668000666 <[email protected]>
Co-authored-by: velvet-shark <[email protected]>
Reviewed-by: @velvet-shark

…aw#91791)

* fix(sandbox): use materialized skill paths in command prompts

* fix(sandbox): resolve backend prompt workdirs

* fix(sandbox): preserve custom backend prompt fallback

---------

Co-authored-by: Vincent Koc <[email protected]>

* Support existing-session browser CDP endpoints

* Fix browser existing-session test fixture type

---------

Co-authored-by: OpenAI Codex <[email protected]>

* fix(browser): validate discovered cdp websocket urls

* fix(browser): validate cdp tab creation websockets

* fix(browser): guard termination cdp websocket

* fix(browser): use .toString() instead of String() to satisfy oxlint no-base-to-string

* fix(browser): avoid cdp termination assertion stringification

* fix(browser): preserve cdp ssrf policy

…1882)

* feat(anthropic): support Claude Fable 5

* test(anthropic): tighten Fable stream fixtures

* fix(anthropic): preserve Vertex input types

* test(anthropic): use provider-ready Vertex effort

* fix(anthropic): support Fable deployment aliases

* fix(anthropic): discard incomplete Fable output

* feat(anthropic): support Fable on Bedrock

* fix(anthropic): preserve Fable reasoning contracts

* refactor(anthropic): unify canonical Claude model policy

* fix(anthropic): satisfy extension thinking types

* test(anthropic): complete canonical alias fixture

* fix(bedrock): scope thinking case declarations

Fix QMD watcher ignore handling for explicitly configured roots whose directory names are normally ignored, and prefer the most-specific configured watch root for overlapping collections.

Validated with focused QMD/tooling tests, full core support boundary tests, green CI, and ClawSweeper re-review.

Verify imported auth profiles and auth runtime state can be read back from the canonical SQLite store before doctor removes legacy JSON auth files.

The migration now keeps legacy JSON in place with a warning when SQLite verification misses expected imported data, avoiding source cleanup before durable readback succeeds.

Verification:
- `scripts/pr review-tests 91740 src/commands/doctor-auth-flat-profiles.test.ts`
- `./node_modules/.bin/oxfmt --check src/commands/doctor-auth-flat-profiles.ts src/commands/doctor-auth-flat-profiles.test.ts`
- `./node_modules/.bin/oxlint src/commands/doctor-auth-flat-profiles.ts src/commands/doctor-auth-flat-profiles.test.ts`
- `scripts/pr-prepare gates 91740` passed `pnpm build` and `pnpm check`; broad local `pnpm test` failed in unrelated baseline shards and was explicitly waived by maintainer.
- Real behavior proof: https://github.com/openclaw/openclaw/actions/runs/27285927515/job/80593035530

Co-authored-by: fuller-stack-dev <[email protected]>
Co-authored-by: Radek Sienkiewicz <[email protected]>

…penclaw#91451)

* fix(mcp): repair OAuth redirect, errors, and unicode schema patterns

Fixes openclaw#91433

Co-authored-by: Cursor <[email protected]>

* fix(mcp): address PR review on redirect compat and OAuth errors

Preserve 127.0.0.1 default, retry DCR with localhost on rejection,
fix lint no-new probes, and test body-less OAuth error responses.

Co-authored-by: Cursor <[email protected]>

* fix(mcp): preserve OAuth redirect and error body text

Persist localhost redirect after DCR fallback so --code exchange matches,
and read body-less foreign response text before SDK error parsing.

Co-authored-by: Cursor <[email protected]>

* fix(mcp): harden OAuth and schema normalization

* fix(mcp): persist only successful OAuth fallback

* fix(mcp): preserve OAuth retry state

---------

Co-authored-by: Cursor <[email protected]>
Co-authored-by: Vincent Koc <[email protected]>

Fix Gateway QMD startup so interval and embedding maintenance are armed when configured, even when the immediate on-boot update is disabled.

Fast-forward path: main -> upstream v2026.6.6, then merge into boon (boon
stays the production source of truth). Consolidates the fork's deltas onto the
new base and drops fork commits upstream has since absorbed.

Target rationale: v2026.6.6 is latest stable and dominates 6.5/6.1 on
thinking-token fixes (PRs openclaw#89440, openclaw#90697, openclaw#91501, openclaw#91149 in 6.5; session
hardening openclaw#91590/openclaw#88630/openclaw#91657 in 6.6) with no regression. Still-open in 6.6:
openclaw#92360 (stale signed thinking-block corruption in persistent named/cron
sessions; fix is post-6.6/beta-only), openclaw#73182, openclaw#92201.

Consolidation decisions:
- KEEP Sentry bundled extension (extensions/sentry-monitor/**): re-applied
  contract entries in bundled-plugin-metadata.test.ts and
  boundary-invariants.test.ts (typed-hook FILES + GUARDS), knip
  rootBundledPluginRuntimeDependencies, @sentry/node root dep. Dropped
  skill-workshop (removed upstream in 6.6).
- KEEP ENG-12929 idle-timeout fix: carried onto renamed
  src/agents/embedded-agent-runner/run/llm-idle-timeout.ts; merged cleanly with
  upstream openclaw#78361/openclaw#77744 per-provider timeout path. 77/77 tests pass.
- KEEP fork CI adaptations: GitHub-hosted runner for codeql-macos (no
  Blacksmith); removed *-testbox.yml workflows (fork has no Testbox infra).
- DROP session-takeover hardening (#2): all picks incl. carried PR openclaw#86584 fence
  fix are now in 6.6 (verified MAX_BENIGN_SESSION_FENCE_* constants present).
- DROP release/perf/qa commits: superseded by 6.6. speech-core removed (gone in
  6.6).

Version set to 2026.6.6-boon.1.

KNOWN GAPS (must run in CI/Testbox; local pnpm blocked by env npm-auth):
- pnpm-lock.yaml needs regeneration: package.json adds @sentry/node and drops
  ajv vs 6.6 lockfile. CI frozen-lockfile check enforces this.
- Full `pnpm build` + full suite not run locally; targeted tests pass
  (idle-timeout 77, sentry contracts 49) via run-vitest.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

# Conflicts:
#	package.json

Reconciles the lockfile with package.json after the v2026.6.6 merge: bundled
Sentry plugin runtime dep @sentry/[email protected] added as a root dependency; ajv
dropped (6.6 no longer imports it). Fixes ERR_PNPM_OUTDATED_LOCKFILE that was
failing every frozen-lockfile CI check on PR #12. Regenerated with pnpm 10.10.0
(--lockfile-only), lockfileVersion 9.0 preserved.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

Root shrinkwrap was stale after adding the @sentry/node bundled-plugin runtime
dep; regenerated via scripts/generate-npm-shrinkwrap.mjs. Fixes check-shrinkwrap
on PR #12. deps:shrinkwrap --all --check now clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

Re-resolved with pnpm 11.2.2 (CI's pinned version). The only change is the
patched-dep entry (@agentclientprotocol/[email protected]) moving from
pnpm 10's verbose hash:/path: form to pnpm 11's compact single-line form. CI's
pnpm 11 install was rewriting the 10.10.0-generated entry, which drifted the
derived npm-shrinkwrap and failed check-shrinkwrap. deps:shrinkwrap --all
--check now clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

Resolves security-fast (pnpm-audit-prod) which flagged 5 high/critical advisories
inherited from upstream 6.6's pinned production deps:
- esbuild 0.28.0 -> 0.28.1 (GHSA-gv7w-rqvm-qjhr; matches upstream 6.7)
- vite 8.0.14 -> 8.0.16 (GHSA-fx2h-pf6j-xcff)
- @vitest/browser 4.1.7 -> 4.1.8 (CRITICAL GHSA-g8mr-85jm-7xhm)
- form-data 2.5.4 -> 2.5.6 (GHSA-hmw2-7cc7-3qxx; existing override updated)
- protobufjs 8.4.0 -> 8.4.1 (GHSA-wcpc-wj8m-hjx6; existing override updated)

vitest family (vitest/@vitest/coverage-v8/@vitest/browser/@vitest/browser-playwright)
kept in lockstep at 4.1.8 so the @vitest/browser bump has a matching peer. Direct
devDeps bumped in package.json; transitive forces via pnpm-workspace.yaml overrides.
Lockfile + all npm-shrinkwrap.json regenerated with pnpm 11.2.2. Audit now clean:
'No high or higher advisories found for production dependencies'.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

6.6's oxlint config is stricter than 5.18; two fork-authored sentry-monitor
sites now fail lint:extensions(+:bundled):
- captures.ts: !!event.deliveryError -> Boolean(...) (no-implicit-coercion)
- dispatch.test.ts: extract call?.[0] to a local before .message access
  (no-unsafe-optional-chaining)

Both files now lint clean (oxlint: 0 errors).

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

The fork runs on GitHub-hosted runners without Blacksmith/Testbox infra. The
testbox workflows (ci-check-testbox, ci-check-arm-testbox, ci-build-artifacts-
testbox, windows-testbox-probe, windows-blacksmith-testbox) were removed in the
merge. Update the 4 upstream test files that read or assert those paths:
- check-workflows.test.ts: drop windows-testbox-probe assertion
- ci-workflow-guards.test.ts: prune 3 testbox paths from workflowPaths array
- package-acceptance-workflow.test.ts: remove CI_CHECK_TESTBOX_WORKFLOW const,
  testboxWorkflow variable, and the Testbox delegation test
- test-projects.test.ts: prune testbox paths from Crabbox/Testbox lane test

197/197 tests pass across the 4 files.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

check-shrinkwrap reports the root npm-shrinkwrap.json stale in CI, but local
regen with CI's node (24.16.0/npm 11.13.0) produces byte-identical output and
--all --check is clean. Add OPENCLAW_SHRINKWRAP_DIFF=1 to print the first 60
differing lines so CI surfaces what it resolves differently (likely a
registry-floated transitive not pinned in pnpm-lock). Revert once diagnosed.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

…gnostic

- hono 4.12.21 -> 4.12.25 resolves newly-published GHSA-88fw-hqm2-52qc (CORS
  middleware reflects any Origin with credentials when origin defaults to
  wildcard; range <4.12.25). Fixes security-fast audit. Lockfile + affected
  npm-shrinkwrap.json regenerated.
- Revert the temporary OPENCLAW_SHRINKWRAP_DIFF diagnostic in ci.yml and
  generate-npm-shrinkwrap.mjs now that the check-shrinkwrap divergence is
  diagnosed (npm hoist nondeterminism in the @sentry/node @opentelemetry/api-logs
  subtree between local and CI runners; needs CI-env regen).

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

check-shrinkwrap fails because local (WSL2) npm hoists @opentelemetry/api-logs
0.56.0 to top-level in the @sentry/node subtree while CI's Linux runner resolves
0.52.1 — npm hoist nondeterminism that's unreproducible locally and has no safe
override (OTel sibling-coupling). Regenerate on ubuntu-latest (CI's environment)
and push the result back so the committed shrinkwrap matches CI resolution.
Temporary; delete after the 6.6 migration lands.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

Diagnostic complete: a clean CI regen (frozen install + deps:shrinkwrap:generate
on ubuntu-latest) reproduces the committed shrinkwrap exactly ('already
current'). check-shrinkwrap's failure is npm hoist nondeterminism across runners
in the @sentry/node @opentelemetry/api-logs subtree (0.52.1/0.56.0/0.218.0 all
legitimately locked), not a content defect. No safe override exists (OTel
sibling version-coupling). Non-blocking: boon has no required checks.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

…ion compare

compareOpenClawReleaseVersions returned null for the fork version scheme
(2026.6.6-boon.1) because parseOpenClawReleaseVersion only matched stable/
alpha/beta/correction forms. Callers then fell back to raw string inequality,
so 'openclaw doctor' judged already-current managed runtime plugins (codex,
acpx) as stale and re-installed/refreshed them on every pass — spurious churn
on every fork host, and the failing doctor tests (openclaw#2048/#2106 in
missing-configured-plugin-install.test.ts) on a -boon.N build.

- npm-registry-spec.ts: strip a trailing -boon.N suffix in
  parseOpenClawReleaseVersion; export stripBoonForkVersionSuffix for callers
  that match raw version strings (companion regexes).
- missing-configured-plugin-install.ts: normalize VERSION through
  stripBoonForkVersionSuffix at the two currentVersion consumption points.
- test: currentOpenClawReleaseBase() also strips -boon.N so derived test
  versions stay valid on fork builds.

All 64 doctor tests + 48 npm-registry-spec tests + update/plugin caller tests
pass. Other compareOpenClawReleaseVersions callers (update-check, plugins/
update, plugins/install) improve identically: fork builds now order correctly
instead of returning null.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>