* fix(agents): answer Claude live control_request can_use_tool via exec policy
Claude CLI emits stream-json control_request frames with subtype
can_use_tool when it wants to use a native tool. The Claude live-session
bridge previously dropped these frames, leaving Claude waiting for a
control_response until the 180/600s no-output timeout fired (see #80819).
Resolve the effective OpenClaw exec policy (per-agent tools.exec -> global
tools.exec -> allowlist/on-miss defaults) once at session-start time and
thread it through fingerprinting and the session record. When a
can_use_tool request arrives:
- Allow native Bash when the resolved policy is security=full, ask=off
(matching the bypassPermissions semantics OpenClaw already documents).
- Otherwise deny with a message that names the resolved policy and
points the agent at OpenClaw MCP tools.
Unsupported control_request subtypes get a structured error response
instead of a silent no-op, and stray control_response frames are
silently dropped. Adds spawn-test coverage for both allow and deny paths.
Fixes#80819
* fix(agents): align Claude live control_request policy with backend defaults
Resolve the effective exec policy through the same defaults that
extensions/anthropic/cli-shared.ts:isOpenClawRequestedYolo and
src/agents/exec-defaults.ts:resolveExecDefaults already use (security
?? "full", ask ?? "off") instead of falling back to a hand-rolled
allowlist/on-miss default that disagreed with the rest of the codebase.
Without this, a default-config OpenClaw deployment launches Claude with
--permission-mode bypassPermissions but the bridge would still deny
Bash control_requests, re-creating the #80819 stall for the very
default-config case the issue reports.
Also thread the effective Claude permission mode into the policy
decision. Prefer the operator's explicit --permission-mode in argv,
falling back to what normalizeClaudePermissionArgs would have inserted
for an un-overridden launch. Native Bash is auto-allowed only when the
effective mode is bypassPermissions AND tools.exec resolves to
full/no-ask, so explicit raw-arg overrides like --permission-mode
default or acceptEdits broaden Claude's native prompting and are
honored by routing through deny.
Adds a no-config regression test (default deployment allows Bash, no
stall) and a permission-mode-override test (tools.exec full/off plus
explicit --permission-mode default in raw args denies). Existing
allow/deny tests continue to pass via the synthesized-mode fallback.
* fix(agents): honor effective exec policy for Claude live Bash
---------
Co-authored-by: Guillaume Thirry <g.thirry@gmail.com>
Summary:
- The PR expands security audit, CLI docs, and tests so `hooks.token` reuse of active Gateway token/password auth is reported while password-mode Gateway startup remains compatible.
- PR surface: Source +178, Tests +311, Docs +14. Total +503 across 14 files.
- Reproducibility: yes. from source inspection: current main forwards a bearer token as both token and passwor ... ecause this review was read-only, but the linked issue and code path make the reproduction high confidence.
Automerge notes:
- PR branch already contained follow-up commit before automerge: fix(cr-fmi-hook-ingress-token-unlocks-password-mode-gateway-auth): ap…
- PR branch already contained follow-up commit before automerge: fix: include trusted proxy password in hooks token reuse check
- PR branch already contained follow-up commit before automerge: fix(gateway): audit hooks password reuse without blocking startup
- PR branch already contained follow-up commit before automerge: fix: Hook ingress token unlocks password-mode gateway auth
Validation:
- ClawSweeper review passed for head 7c796b22ec.
- Required merge gates passed before the squash merge.
Prepared head SHA: 7c796b22ec
Review: https://github.com/openclaw/openclaw/pull/86453#issuecomment-4533831028
Co-authored-by: Coy Geek <65363919+coygeek@users.noreply.github.com>
Co-authored-by: jesse-merhi <79823012+jesse-merhi@users.noreply.github.com>
Co-authored-by: clawsweeper <274271284+clawsweeper[bot]@users.noreply.github.com>
Co-authored-by: clawsweeper[bot] <274271284+clawsweeper[bot]@users.noreply.github.com>
Approved-by: jesse-merhi
Move the plain-text tool-call promotion wrapper out of the public provider stream SDK helper and into a private local-only bundled-provider runtime seam.
* feat(imessage): support thumb approval reactions
Mirrors openclaw#85477 (WhatsApp) for the iMessage channel. iMessage can now
deliver exec/plugin approval prompts via the existing imsg/BlueBubbles
transport and resolve approvals from 👍 (allow-once) / 👎 (deny) tapbacks.
Allow-always remains on the manual /approve <id> allow-always fallback.
What changed:
- New approval surfaces under extensions/imessage/src/:
approval-auth.ts, approval-resolver.ts, approval-reactions.ts,
approval-handler.runtime.ts, approval-native.ts (+ tests for each).
- channel.ts wires base.approvalCapability to the new iMessage capability.
- send.ts appends the 👍/👎 hint to outbound /approve prompts and registers
the reaction binding (keyed by accountId + chat_guid/chat_identifier/
chat_id/handle + messageId) after a successful send.
- monitor/monitor-provider.ts resolves approval reactions ahead of the
normal inbound decision pipeline so resolution bypasses
reactionNotifications gating and runs its own actor authorization.
- runtime.ts now exports getIMessageRuntime / getOptionalIMessageRuntime so
approval-reactions can open a persistent keyed store for binding state
across gateway restarts.
What did NOT change:
- Core approval surfaces in src/gateway/server-methods/* and src/infra/*
remain channel-agnostic; the channels.imessage.allowFrom field already
exists and is reused as the approver list for reactions.
- Other channels and the manual /approve sender-authorized path are
untouched.
* fix(imessage): address codex review findings on thumb approvals
Addresses 15 findings from the multi-angle codex review:
Critical (correctness / blocking):
- Register CHANNEL_APPROVAL_NATIVE_RUNTIME_CONTEXT_CAPABILITY in the iMessage
monitor so the gateway can actually deliver native approval prompts via
approval-handler.runtime.ts (it was dead code without the context lease).
- DM tapback approvals never resolved because send keyed by handle while
inbound preferred chat_guid. Register and look up under EVERY available
conversation key (chat_guid / chat_identifier / chat_id / handle); inbound
probes them all and accepts the first hit.
- Reaction binding now requires the bridge's GUID string (rejecting numeric
ROWIDs) so the binding key matches inbound reacted_to_guid.
- Outbound regex now requires both a canonical `ID: <approvalId>` header AND
a matching `/approve <id> <decision>` line, so non-approval messages that
legitimately mention /approve syntax no longer get a phantom reaction
binding (and can no longer resolve a colliding live approval).
- Drop is_from_me reaction events so cross-device echoes of the operator's
own tap cannot self-approve when their handle is in allowFrom.
High (operability / cleanup):
- Non-ApprovalNotFound errors now log at warn via the runtime child logger
(no longer hidden behind OPENCLAW_LOG_LEVEL=debug).
- In-memory binding is cleared on successful resolve so a toggle 👍→👎 (or
chat.db replay) does not refire and emit a misleading 'expired approval'
log line. Removed tapbacks are also owned by the shortcut and not surfaced
as noisy reaction system events.
- Move resolveIMessageReactionContext (and its helpers) to a slim
monitor/reaction-context.ts so approval-reactions.ts no longer transitively
pulls monitor/inbound-processing.ts (14+ heavy runtime modules) into the
hot channel.ts entrypoint per extensions/CLAUDE.md.
Medium (consistency / future-proofing):
- Native runtime exec pending payload now passes agentId, ask, and
sessionKey through buildExecApprovalPendingReplyPayload so the two
delivery routes produce identical operator-visible prompts.
- Both delivery paths now use addIMessageApprovalReactionHintToText (single
insertion point after ID:) so the hint cannot be double-emitted by the
native runtime path bypassing the idempotency guard.
- Extract replaceApprovalIdPlaceholder into a shared approval-text.ts that
escapes `$` in the replacement string so an approvalId containing
`$&`/`$1`-`$9`/`$$` cannot interpolate into the outbound text.
- In-memory Map now stores TTL alongside each entry and prunes expired
bindings on each register so the gateway no longer accumulates an
unbounded reaction-target Map.
- bindPending refuses to bind when accountId is missing or the approval is
already expired, with explicit error logs instead of silent no-ops.
- Reject chat_id=0 as a synthetic key value (chat.db ROWIDs start at 1).
- Drop dead getIMessageRuntime export — only the optional accessor is used.
Documentation:
- docs/channels/imessage.md gains an 'Approval reactions (👍 / 👎)' accordion
documenting the reaction emoji map, allowFrom approver requirement, the
/approve <id> allow-always manual fallback, and the deliberate change to
/approve command authorization for users with non-empty allowFrom.
- CHANGELOG.md entry added under 2026.5.24.
Tests: 411 iMessage tests pass (was 406). Added explicit coverage for the
DM key-mismatch fix, the regex-tightening fix, the is_from_me guard, the
clear-on-success behavior, and the approval-id `$` escape.
* test(imessage): match WhatsApp approval-native test coverage
Backfills the nine cases from extensions/whatsapp/src/approval-native.test.ts
that weren't mirrored in iMessage:
- target-mode exec + plugin prompt rendering with the canonical hint
- target-mode availability when no iMessage target matches
- agentFilter / sessionFilter applied to native handling
- account-scoped target enabled/disabled per account
- shouldSuppressForwardingFallback session-origin exact-match cases
- shouldSuppressForwardingFallback off when native cannot bind (locks down
the targets-only forwarding path the Lobster live deploy exercised)
- both-mode explicit + unscoped target suppression
- group-origin tapback approvals require explicit approvers
Tests: extensions/imessage/src/approval-native.test.ts 21 passed (was 11).
Total iMessage approval-specific cases now 49 (was 40).
* fix(imessage): preserve service-prefixed direct handles as approvers
ClawSweeper P1 review finding on #85952. normalizeIMessageApproverId was
calling looksLikeIMessageExplicitTargetId() to reject conversation-target
prefixes, but that helper also matches the imessage:/sms:/auto: service
prefixes — which are valid direct-handle forms. Any allowFrom entry like
'imessage:+15551230000' dropped to undefined, leaving approvers empty,
which:
- silently denied reaction resolution ('reactions require explicit
approvers'), and
- let text /approve fall back to implicit same-chat authorization.
Fix: normalize first via normalizeIMessageHandle (strips the service
prefix), then reject only chat_id:/chat_guid:/chat_identifier:
conversation-target shapes that remain after normalization.
Tests:
- approval-auth.test.ts: assert the resolved approver list contains the
normalized handle, plus the corollary that a non-matching sender is
explicitly rejected (no longer masked by the implicit-same-chat
fallback). Add a separate case covering chat_id/chat_guid/
chat_identifier rejection (with and without a service prefix).
- approval-reactions.test.ts: reaction resolution end-to-end with a
service-prefixed allowFrom entry — proves resolveIMessageApproval is
called rather than silently denied.
Focused suite: 48 passed (was 47).
* test(imessage): satisfy strict buildPendingPayload signature in render tests
CI check:test-types caught that the render.exec/render.plugin
buildPendingPayload calls were passing accountId (not in the type
signature). The signature is { cfg, request, target, nowMs }. Replace
accountId with target on the four render-test sites so the strict
test-types pass matches the SDK contract:
- it('renders thumbs-only reaction hints in exec approval prompts')
- it('renders thumbs-only reaction hints in plugin approval prompts ...')
- it('renders target-mode exec prompts with concrete thumbs-only ...')
- it('renders target-mode plugin prompts with concrete thumbs-only ...')
Verified locally with pnpm check:test-types (tsgo:core:test +
tsgo:extensions:test). 49 approval-specific tests still pass.
* fix(imessage): probe every tapback GUID form for approval lookup
ClawSweeper P1 review finding on #85952. readApprovalReactionEvent was
only using reaction.targetGuid (the first/normalized form), but
resolveIMessageReactionContext produces reaction.targetGuids = [normalized,
raw] for both `abc-123` and `p:0/abc-123` forms. If the imsg bridge
returned 'p:0/<guid>' from send() and send.ts registered the binding under
that prefixed key, the inbound resolver probing only the unprefixed form
would miss and the tapback would silently fall through.
Fix:
- Surface every GUID candidate in IMessageApprovalReactionEvent
(messageIdCandidates).
- maybeResolveIMessageApprovalReaction now probes each candidate in
precedence order; first hit wins.
- On success / ApprovalNotFoundError, clear the binding under all
candidate keys so toggle/replay does not refire.
Tests: extensions/imessage/src/approval-reactions.test.ts gains a
'resolves a reaction when the binding was registered under a p:0/…
prefixed GUID and the tapback surfaces both forms' regression case;
22/22 reaction tests pass. Full iMessage suite: 424/424.
* fix(imessage): native approval binding requires GUID, not numeric id
ClawSweeper third P1 review finding on #85952. approval-handler.runtime.ts
deliverPending was using result.messageId as the approval-reaction binding
key, but that field can be a numeric ROWID coerced to a string ('12345')
when the imsg bridge returns only message_id. Inbound tapbacks carry
reacted_to_guid which is always a GUID, so a numeric-id binding can never
match.
Fix mirrors the send.ts forwarding-path treatment:
- IMessageSendResult now exposes a separate guid?: string field, populated
from the same resolveOutboundMessageGuid helper send.ts already uses for
the forwarding-path binding. The generic messageId field is unchanged so
reply-cache, echo-cache, and receipt-building paths still see the
broadest id form.
- deliverPending now binds against result.guid; when it's undefined (numeric
ROWID or 'ok'/'unknown' placeholders), the function returns null instead
of binding against an id the inbound tapback can't possibly match.
Tests: approval-handler.runtime.test.ts gets a deliverPending GUID-only
binding describe block with three regression cases (numeric ROWID refused,
GUID accepted, ok/unknown placeholders refused). vi.mock isolates
sendMessageIMessage so the cases run synchronously without spawning imsg.
11 tests pass across handler.runtime + send specs.
---------
Co-authored-by: Omar Shahine <10343873+omarshahine@users.noreply.github.com>
* codex: honor verbose in group dispatch
* codex: address group verbose review findings
Record the final local review pass for the group /verbose PR.
Codex review against origin/main completed clean after tightening the shared group progress gate, keeping public plugin hook types stable, preserving ACP hidden tool boundaries, and adding regressions for live verbose gating and progress-callback suppression.
* codex: require explicit group verbose progress
Normal group tool/progress summaries now require an explicit session verbose override instead of inherited agent verbose defaults.
This addresses the PR review concern that existing verboseDefault configurations could expose group progress after upgrade. DMs and forum-topic behavior continue to use the effective verbose state, while normal groups use the live explicit session verbose state set by /verbose on|full|off.
* codex: document Slack group verbose caveat
* fix(channels): simplify verbose progress gating
* docs(changelog): note verbose channel fix
* fix(channels): preserve quiet default for group progress
* fix(channels): keep verbose error policy dynamic
* fix(channels): default verbose progress off everywhere
* fix(channels): keep followup verbose default quiet
* fix(channels): latch visible tool-error progress
* fix(channels): track failed verbose progress events
* fix(channels): latch delivered tool errors
* fix(channels): prevent progress opt-out bypass
* fix(channels): isolate followup error warning state
* fix(channels): keep full verbose followup warnings
* fix(channels): latch tool errors after visible progress
* fix(channels): require visible followup failure progress
* fix(channels): refresh followup verbose state
* fix(channels): honor live verbose for error details
* test(channels): expect live verbose off warning mode
* fix(channels): preserve static tool error suppression semantics
* fix(channels): bypass acp for colon verbose commands
* fix(channels): narrow dynamic tool warning override
* fix(channels): gate compaction notices on live verbose
* fix(channels): suppress quiet followup compaction callbacks
* fix(channels): suppress tts for hidden tool summaries
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* fix: harden package URL downloads
Guard package acceptance URL downloads with HTTPS-only validation, no embedded credentials, private/special-use DNS and IP rejection, manual redirect checks, bounded timeout/size limits, pinned lookup, and atomic temp-file writes. Add tooling tests for unsafe URLs, redirect validation, size limits, and successful writes.
* fix: cancel redirect response bodies before closing dispatcher
ClawSweeper P2: the redirect branch in openPackageDownloadResponse cleared
the timeout and awaited dispatcher.close() without first cancelling
response.body. Undici's close() is graceful — it waits for in-flight
requests to complete — so a malicious redirect with a slow/never-ending
body could hang the hardened downloader.
Fix: call response.body?.cancel() before dispatcher.close() to abort the
redirect body immediately.
Test: add a regression test that uses a ReadableStream with an indefinite
interval to simulate a hanging body, and asserts cancel() was called.
Refs: clawsweeper review on PR #85512
* test: harden redirect body cancellation race in regression test
Guard the ReadableStream controller.enqueue() call with a cancelled
flag and try/catch to prevent ERR_INVALID_STATE when the interval
fires after cancel() closes the controller.
* fix: cancel final response body before closing dispatcher in downloadUrl
ClawSweeper P2: the HTTP-error and declared-oversize early-exit paths
in downloadUrl threw before consuming or canceling response.body. The
finally block then cleared the timeout and awaited graceful
dispatcher.close() with the body still open, allowing a slow/never-ending
response to hang release tooling.
Fix: add response.body?.cancel() in the finally block before
dispatcher.close().
Tests: add two regressions:
- HTTP 500 with slow body: asserts cancel() called before dispatcher close
- Declared content-length oversize with slow body: same assertion
* fix: add trusted package URL source policy
* fix: keep package URL resolver dependency-free
* test: cover encoded IPv6 package URL bypasses
* docs: sync package acceptance source overview
* docs: restore release doc formatting
* docs: sync package acceptance trusted-url source
* test: cover dotted IPv4 embedded IPv6 package URLs
* fix: parse dotted IPv4 embedded in IPv6 package URLs
* test: isolate anthropic pruning defaults
* test: move anthropic dated model coverage
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* fix(status): show configured cost for aws-sdk models
Decouple status cost display from provider auth mode so explicit model pricing is used for Bedrock and other non-api-key providers. Include cache read/write tokens in the status cost estimate and cover the behavior with regression tests.
* fix: show configured response usage costs
* docs: align configured cost visibility
* fix(status): keep usage tokens mode cost-free
---------
Co-authored-by: ItsOtherMauridian <165866613+ItsOtherMauridian@users.noreply.github.com>
Co-authored-by: ItsOtherMauridian <itsothermauridian@users.noreply.github.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
