fix(agents): honor configured CLI resume timeouts by ai-hpc · Pull Request #90912 · openclaw/openclaw

ai-hpc · 2026-06-06T11:10:05Z

Summary

Honor configured agent/default timeouts when resolving the no-output watchdog for resumed CLI-backed embedded runs.
Plumb the existing explicit run-timeout signal through CLI dispatch so per-run timeouts keep the same behavior as cron.
Keep inherited user resumed sessions on the existing 180s watchdog ceiling when no timeout intent is configured.

Real behavior proof

behavior: A resumed CLI-backed embedded run with a deliberate 600s timeout should not keep the default 180s no-output ceiling. The watchdog should scale from the configured run timeout, while ordinary inherited user resumes keep the existing 180s guard.

environment: Local source checkout rebased on 0a08625d79, Node v22.22.2, pnpm 11.2.2, branch ai-hpc/90764-embedded-timeout.

steps:

Checked current github/main source in src/agents/cli-runner/reliability.ts: resumed CLI runs only switch to fresh watchdog defaults when trigger === "cron", so user-triggered resumed runs stay on CLI_RESUME_WATCHDOG_DEFAULTS.maxMs = 180000.
Applied the fix and ran a resolver probe for a 600s resumed CLI run.
Added dispatch-level coverage proving timeout: "600" reaches runAgentAttempt as runTimeoutOverrideMs: 600000.
Ran focused CLI reliability tests, the touched agent-command dispatch test, the attempt-execution CLI test, and the core type gate.

evidence:

{
  "inherited": 180000,
  "configured": 480000,
  "explicitRun": 480000,
  "cron": 480000
}

observedResult: After the fix, configured/default and explicit per-run 600s resumed CLI runs resolve the no-output watchdog to 480s. A plain inherited user resume still resolves to 180s, preserving the default guard.

notTested: I did not replay the direct custom-api Anthropic run from #90764. This PR covers the confirmed CLI-backed resumed-session watchdog branch only. I also did not run a live long-running CLI provider session because the resolver path and CLI reliability tests cover the timeout calculation without waiting 180s.

Verification

OPENCLAW_VITEST_MAX_WORKERS=1 pnpm test src/agents/cli-runner.reliability.test.ts
OPENCLAW_VITEST_MAX_WORKERS=1 pnpm test src/agents/cli-runner.reliability.test.ts -- -t resolveCliNoOutputTimeoutMs
OPENCLAW_VITEST_MAX_WORKERS=1 pnpm test src/agents/agent-command.live-model-switch.test.ts -- -t "passes explicit timeout overrides"
OPENCLAW_VITEST_MAX_WORKERS=1 pnpm test src/agents/agent-command.live-model-switch.test.ts
OPENCLAW_VITEST_MAX_WORKERS=1 pnpm test src/agents/command/attempt-execution.cli.test.ts
pnpm tsgo:core:test
OPENCLAW_CHECK_CHANGED_REMOTE_CHILD=1 pnpm check:changed --base github/main --head HEAD

Related: #90764

clawsweeper · 2026-06-06T11:11:40Z

Codex review: needs maintainer review before merge. Reviewed June 10, 2026, 9:58 AM ET / 13:58 UTC.

Summary
The PR propagates explicit CLI run-timeout intent and configured agent timeout presence into resumed-run watchdog selection while preserving the 180-second ceiling for inherited user resumes.

PR surface: Source +30, Tests +53. Total +83 across 8 files.

Reproducibility: yes. at source level: current main applies the 180-second resumed watchdog despite deliberate longer timeout intent, and the focused resolver path clearly demonstrates that mismatch. The broader custom Anthropic scenario has not been reproduced.

Review metrics: none identified.

Merge readiness
Overall: 🐚 platinum hermit
Proof: 🌊 off-meta tidepool
Patch quality: 🐚 platinum hermit
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

none.

Risk before merge

[P2] For resumed CLI runs with a deliberate or configured long timeout, a genuinely stalled process can now remain alive substantially longer than the previous 180-second ceiling before OpenClaw terminates it.
[P2] The linked report at [2026.6.2-beta.1] Embedded agent runs abort at ~180s wall-clock regardless of agent/provider timeoutSeconds #90764 used a custom Anthropic embedded path that was not replayed; this patch should not be treated as proof that every timeout cause described there is resolved.

Maintainer options:

Land the scoped CLI fix (recommended)
Accept the longer configured watchdog window for resumed CLI runs and keep the broader linked report open until its non-CLI path is independently verified.
Preserve the fixed resume ceiling
Pause or close this PR if maintainers want the 180-second no-output limit to override deliberate longer run timeouts for all user-triggered resumed CLI sessions.

Next step before merge

[P2] No concrete mechanical defect needs repair; a maintainer should explicitly accept or reject the longer stalled-process retention window and then land or close this member-authored PR.

Security
Cleared: The focused patch only changes internal timeout propagation, watchdog selection, and tests; it introduces no dependency, workflow, permission, secret, or supply-chain changes.

Review details

Best possible solution:

Land the narrow timeout-intent correction if maintainers accept configured run timeouts as authoritative for resumed CLI watchdog scaling, while keeping the broader linked issue open for independent reproduction of its custom-provider path.

Do we have a high-confidence way to reproduce the issue?

Yes at source level: current main applies the 180-second resumed watchdog despite deliberate longer timeout intent, and the focused resolver path clearly demonstrates that mismatch. The broader custom Anthropic scenario has not been reproduced.

Is this the best way to solve the issue?

Yes for the confirmed CLI-backed path: carrying explicit timeout intent is narrower and more maintainable than inferring intent from resolved values, and it preserves backend overrides and inherited resume behavior.

AGENTS.md: found and applied where relevant.

Codex review notes: reasoning high; reviewed against bb6e47729cf8.

Label changes

Label justifications:

P2: This is a focused reliability correction for a bounded resumed CLI agent path with limited blast radius.
merge-risk: 🚨 availability: Merging intentionally extends how long configured resumed CLI runs may remain silent before process termination.
rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🌊 off-meta tidepool and patch quality is 🐚 platinum hermit.
status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Not applicable: The author has repository member status, and the write-access exception permits the focused resolver output and local tests to serve as recorded proof without an external-contributor live-proof gate.

Evidence reviewed

PR surface:

Source +30, Tests +53. Total +83 across 8 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	6	38	8	+30
Tests	2	54	1	+53
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	8	92	9	+83

What I checked:

Current-main mismatch: Current main selects fresh watchdog defaults for resumed runs only when the trigger is cron, so user-triggered resumes retain the 180-second maximum despite deliberate longer timeout intent. (src/agents/cli-runner/reliability.ts:27, bb6e47729cf8)
Focused implementation: The PR carries a separate timeout-intent value through agent command dispatch, attempt execution, and CLI execution rather than inferring intent from the resolved deadline. (src/agents/agent-command.ts:659, 4f48041ddb77)
Timeout semantics: Explicit zero is normalized to the maximum timer-safe timeout, confirming that a separate intent signal is safer than comparing resolved timeout values with configured defaults. (src/agents/timeout.ts:23, bb6e47729cf8)
Regression coverage: The PR separately covers explicit run timeout, configured default timeout, cron timeout, backend watchdog override, and inherited resumed-session behavior. (src/agents/cli-runner.reliability.test.ts:2006, 4f48041ddb77)
Codex resume contract: Current Codex supports the non-interactive exec --json resume form used by OpenClaw. (../codex/codex-rs/cli/src/main.rs:2610, d2f6d23c6cb0)
Codex activity output: Codex emits each mapped JSON thread event to stdout while processing a turn, giving OpenClaw observable activity with which to reset its no-output watchdog. (../codex/codex-rs/exec/src/event_processor_with_jsonl_output.rs:104, d2f6d23c6cb0)

Likely related people:

Vincent Koc: Current-main blame and file history associate the CLI watchdog helper and its fresh/resume bounds with this contributor's commit. (role: introduced behavior; confidence: high; commits: 52bc2a12bc8c; files: src/agents/cli-runner/reliability.ts, src/agents/cli-watchdog-defaults.ts, src/agents/cli-runner/execute.ts)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

vincentkoc · 2026-06-11T01:05:41Z

Maintainer verification complete.

The original change had the right intent but timeout provenance was incomplete across normal channel runs, queued follow-ups, heartbeats, and CLI-backed subagents. I moved the decision to the shared CLI boundary, carried explicit reply timeout provenance through the queue/fallback paths, and preserved the shorter inherited resume watchdog for subagents.

Focused proof: node scripts/run-vitest.mjs src/agents/agent-command.live-model-switch.test.ts src/agents/cli-runner.reliability.test.ts src/auto-reply/reply/agent-runner-execution.test.ts src/auto-reply/reply/followup-runner.test.ts — 354 passed.
Fresh autoreview after all fixes: clean; no accepted/actionable findings.

Best-fix judgment: landable. The final shape distinguishes deliberate/configured run timeouts from inherited defaults at the shared CLI owner boundary instead of guessing from the final numeric timeout.

* fix(agents): honor configured CLI resume timeouts * fix(agents): pass explicit CLI timeout overrides * fix(agents): preserve timeout provenance * test(agents): cover configured timeout provenance * fix(agents): resolve CLI timeout provenance centrally * fix(agents): carry reply timeout provenance * fix(agents): forward CLI run lane --------- Co-authored-by: Vincent Koc <[email protected]>

openclaw-barnacle Bot added agents Agent runtime and tooling size: S labels Jun 6, 2026

ai-hpc force-pushed the ai-hpc/90764-embedded-timeout branch from 4ccca30 to 86df1c3 Compare June 6, 2026 12:41

ai-hpc force-pushed the ai-hpc/90764-embedded-timeout branch from 86df1c3 to 5069567 Compare June 10, 2026 12:49

ai-hpc force-pushed the ai-hpc/90764-embedded-timeout branch 3 times, most recently from bb915da to a10c9c1 Compare June 11, 2026 00:08

vincentkoc self-assigned this Jun 11, 2026

ai-hpc and others added 5 commits June 11, 2026 10:01

fix(agents): honor configured CLI resume timeouts

11a1d07

fix(agents): pass explicit CLI timeout overrides

b3cacfc

fix(agents): preserve timeout provenance

bb8bf9c

test(agents): cover configured timeout provenance

7d7dc7a

fix(agents): resolve CLI timeout provenance centrally

a1bc89e

vincentkoc added 2 commits June 11, 2026 10:01

fix(agents): carry reply timeout provenance

e9c6abf

fix(agents): forward CLI run lane

b7b061f

vincentkoc force-pushed the ai-hpc/90764-embedded-timeout branch from a10c9c1 to b7b061f Compare June 11, 2026 01:05

vincentkoc merged commit 0328ca5 into openclaw:main Jun 11, 2026
22 checks passed

Haderach-Ram mentioned this pull request Jun 11, 2026

Ecosystem Digest — 2026-06-11 Haderach-Ram/openclaw-radar#36

Open

wangwllu mentioned this pull request Jun 12, 2026

fix: handle Claude CLI synthetic placeholders #90799

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(agents): honor configured CLI resume timeouts#90912

fix(agents): honor configured CLI resume timeouts#90912
vincentkoc merged 7 commits into
openclaw:mainfrom
ai-hpc:ai-hpc/90764-embedded-timeout

ai-hpc commented Jun 6, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 6, 2026 •

edited

Loading

Uh oh!

vincentkoc commented Jun 11, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

ai-hpc commented Jun 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Real behavior proof

Verification

Uh oh!

clawsweeper Bot commented Jun 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vincentkoc commented Jun 11, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

ai-hpc commented Jun 6, 2026 •

edited

Loading

clawsweeper Bot commented Jun 6, 2026 •

edited

Loading