fix(loop-detection): escalate generic_repeat to critical at criticalThreshold by jwchmodx · Pull Request #60248 · openclaw/openclaw

jwchmodx · 2026-04-03T11:17:16Z

Summary

The generic_repeat detector only checked warningThreshold and always returned level: "warning", making criticalThreshold a no-op for the most common runaway loop pattern (same tool, same arguments, repeated calls).

Added a criticalThreshold check in the generic_repeat code path, before the existing warningThreshold check
Updated the test that was asserting the broken warn-only behavior to assert the correct critical escalation

This makes generic_repeat consistent with known_poll_no_progress and ping_pong, which both escalate to "critical" at criticalThreshold.

Detector	warningThreshold	criticalThreshold
`known_poll_no_progress`	⚠️ Warns	✅ Blocks
`ping_pong`	⚠️ Warns	✅ Blocks (with noProgressEvidence)
`generic_repeat` (before)	⚠️ Warns	❌ Not checked
`generic_repeat` (after)	⚠️ Warns	✅ Blocks

Test plan

Existing 29 tests pass
Updated test now asserts level: "critical" and detector: "generic_repeat" at criticalThreshold

🤖 Generated with Claude Code

greptile-apps · 2026-04-03T11:19:18Z

Greptile Summary

This PR correctly fixes the generic_repeat loop detector so it escalates to "critical" at criticalThreshold (previously it only ever returned "warning"), making it consistent with the known_poll_no_progress and ping_pong detectors. It also bundles an unrelated fix in extensions/mattermost that prevents DM replies from being threaded by adding a required kind: ChatType guard to resolveMattermostReplyRootId; all three call sites are updated and new tests cover the DM-gating behavior.

Confidence Score: 5/5

Safe to merge — logic is correct, all call sites are updated, and only P2 suggestions remain.

No P0 or P1 issues found. The critical-threshold check for generic_repeat is correctly ordered before the warning check, matching the established pattern. The mattermost DM threading guard is complete and tested. Remaining comments are style suggestions only.

No files require special attention.

Comments Outside Diff (1)

extensions/mattermost/src/mattermost/monitor.ts, line 160-173 (link)

Unrelated fix bundled without PR description coverage

The kind: ChatType parameter addition to resolveMattermostReplyRootId — preventing DM threading — is a separate behaviour fix not mentioned in the PR description or summary. Since the function is exported, reviewers need to know it was intentionally included and that all three call sites (lines 536, 744, 1456) were updated. The change itself is correct, but it would be worth adding a note in the PR description so it can be tracked against the issue it resolves.

Prompt To Fix With AI

This is a comment left during a code review.
Path: extensions/mattermost/src/mattermost/monitor.ts
Line: 160-173

Comment:
**Unrelated fix bundled without PR description coverage**

The `kind: ChatType` parameter addition to `resolveMattermostReplyRootId` — preventing DM threading — is a separate behaviour fix not mentioned in the PR description or summary. Since the function is `export`ed, reviewers need to know it was intentionally included and that all three call sites (lines 536, 744, 1456) were updated. The change itself is correct, but it would be worth adding a note in the PR description so it can be tracked against the issue it resolves.

How can I resolve this? If you propose a fix, please make it concise.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

Prompt To Fix All With AI

This is a comment left during a code review.
Path: src/agents/tool-loop-detection.test.ts
Line: 302-316

Comment:
**Missing count assertion in updated test**

The renamed test now asserts `level: "critical"` and `detector: "generic_repeat"`, but the previous test also checked `result.count`. Adding `expect(loopResult.count).toBe(CRITICAL_THRESHOLD)` would keep coverage parity with the existing warning-level test (line 296) and pin the count value returned by the new critical path.

```suggestion
      expect(loopResult.stuck).toBe(true);
      if (loopResult.stuck) {
        expect(loopResult.level).toBe("critical");
        expect(loopResult.detector).toBe("generic_repeat");
        expect(loopResult.count).toBe(CRITICAL_THRESHOLD);
        expect(loopResult.message).toContain("CRITICAL");
      }
```

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: extensions/mattermost/src/mattermost/monitor.ts
Line: 160-173

Comment:
**Unrelated fix bundled without PR description coverage**

The `kind: ChatType` parameter addition to `resolveMattermostReplyRootId` — preventing DM threading — is a separate behaviour fix not mentioned in the PR description or summary. Since the function is `export`ed, reviewers need to know it was intentionally included and that all three call sites (lines 536, 744, 1456) were updated. The change itself is correct, but it would be worth adding a note in the PR description so it can be tracked against the issue it resolves.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "fix(loop-detection): escalate generic_re..." | Re-trigger Greptile}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e02478669a

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9a02183ce7

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

…hreshold The generic_repeat detector only checked warningThreshold and always returned level "warning", making criticalThreshold effectively a no-op for the most common runaway loop pattern (same tool + same args). Add a critical-threshold check before the warning check, consistent with how known_poll_no_progress and ping_pong already behave. Fixes openclaw#60111 Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

vincentkoc · 2026-04-28T07:37:39Z

ProjectClownfish pushed a narrow repair to this branch so the original contributor path can stay canonical.

Source PR: #60248
Validation: pnpm test src/agents/tool-loop-detection.test.ts -- --reporter=dot; pnpm check:changed
Contributor credit is preserved in the branch history and PR context.

clawsweeper · 2026-04-28T07:42:05Z

Thanks for the context here. I did a careful shell check against current main, and this is already implemented.

Current main already contains the complete no-progress-gated generic_repeat critical escalation, with unit, before-tool-call e2e, diagnostic, docs, changelog, and release-note coverage from the merged replacement PR.

So I’m closing this as already implemented rather than keeping a duplicate issue open.

Review details

Best possible solution:

Keep the merged current-main implementation from #80668 and close this superseded contributor branch instead of merging a narrower stale diff.

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

Yes for the original bug: the detector/test path around repeated same-tool calls showed the warning-only gap before the fix. On current main, that failure no longer reproduces because the source and e2e tests now cover generic_repeat critical blocking for identical outcomes.

Is this the best way to solve the issue?

Yes. The current-main implementation is the narrow maintainable solution because it blocks only repeated no-progress outcomes and includes the before-tool-call e2e/diagnostic coverage this PR was missing.

Security review:

Security review cleared: The reviewed PR and merged replacement only affect loop-detection behavior, tests, docs, and changelog, with no dependency, CI, secret, install, or package execution changes.

What I checked:

current_main_source: Current main gates the generic_repeat critical path on noProgressStreak >= resolvedConfig.criticalThreshold and returns level: "critical" with detector: "generic_repeat". (src/agents/tool-loop-detection.ts:587, 3e425dde0d62)
current_main_tests: Current main has direct detector coverage for blocking generic no-progress loops at the critical threshold and before-tool-call e2e coverage for blocking plus warning-to-critical diagnostics. (src/agents/pi-tools.before-tool-call.e2e.test.ts:286, 3e425dde0d62)
current_main_docs: The loop-detection docs now state that criticalThreshold blocks repetitive no-progress patterns and that genericRepeat blocks only when the same calls return identical outcomes. Public docs: docs/tools/loop-detection.md. (docs/tools/loop-detection.md:80, 3e425dde0d62)
merged_fix_pr: Merged PR fix: abort generic no-progress tool loops #80668 implemented the same central behavior and added the missing before-tool-call e2e and diagnostic coverage; its merge commit is the fix provenance. (src/agents/tool-loop-detection.ts:609, 678b2510b22c)
release_provenance: Release v2026.5.12-beta.1 includes the note: "Agents: abort generic repeated no-progress tool loops at the critical threshold when identical calls keep returning identical outcomes. (fix: abort generic no-progress tool loops #80668)"; v2026.5.10-beta.5 did not include that line. (678b2510b22c)
related_search: Canonical search found this PR plus merged replacement fix: abort generic no-progress tool loops #80668 and related loop-detector reports, confirming the remaining central work is already represented by the merged fix. (3e425dde0d62)

Likely related people:

frankekn: Merged PR fix: abort generic no-progress tool loops #80668 implemented the no-progress-gated generic repeat block and touched the central detector, tests, e2e, docs, and changelog paths. (role: current fix author and merger; confidence: high; commits: cffba00372f3, 7fa287cd0f88, 678b2510b22c; files: src/agents/tool-loop-detection.ts, src/agents/tool-loop-detection.test.ts, src/agents/pi-tools.before-tool-call.e2e.test.ts)
steipete: Repository history for commit 076df94 identifies the original configurable tool loop detection work that created this detector/config surface. (role: introduced behavior; confidence: medium; commits: 076df941a326; files: src/agents/tool-loop-detection.ts, docs/tools/loop-detection.md)
vincentkoc: The PR discussion records a narrow repair pushed to this branch to keep the contributor path canonical, changing the generic repeat behavior to require no-progress evidence. (role: branch repair contributor; confidence: medium; commits: 9fd4b85d9a4b; files: src/agents/tool-loop-detection.ts, src/agents/tool-loop-detection.test.ts, docs/tools/loop-detection.md)

Codex review notes: model gpt-5.5, reasoning high; reviewed against 3e425dde0d62; fix evidence: release v2026.5.12-beta.1, commit 678b2510b22c.

greptile-apps Bot reviewed Apr 3, 2026

View reviewed changes

Comment thread src/agents/tool-loop-detection.test.ts

chatgpt-codex-connector Bot reviewed Apr 3, 2026

View reviewed changes

Comment thread src/agents/tool-loop-detection.ts

chatgpt-codex-connector Bot reviewed Apr 3, 2026

View reviewed changes

Comment thread src/infra/provider-usage.fetch.minimax.ts

steipete mentioned this pull request Apr 3, 2026

fix(xiaomi): disable thinking for MiMo reasoning models #60283

Closed

jwchmodx and others added 2 commits April 28, 2026 07:37

fix(loop-detection): block generic repeats with no-progress evidence

9fd4b85

vincentkoc force-pushed the fix/generic-repeat-critical-threshold branch from 9a02183 to 9fd4b85 Compare April 28, 2026 07:37

vincentkoc added clownfish:human-review labels Apr 28, 2026

openclaw-barnacle Bot added docs Improvements or additions to documentation agents Agent runtime and tooling size: S labels Apr 28, 2026

vincentkoc added clawsweeper Tracked by ClawSweeper automation and removed clownfish:merge-ready labels Apr 28, 2026

clawsweeper Bot mentioned this pull request May 11, 2026

[Bug]: Openclaw is stupid again: 200000 tool calls, no feedback, interruptions #48947

Closed

clawsweeper Bot closed this May 12, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(loop-detection): escalate generic_repeat to critical at criticalThreshold#60248

fix(loop-detection): escalate generic_repeat to critical at criticalThreshold#60248
jwchmodx wants to merge 2 commits into
openclaw:mainfrom
jwchmodx:fix/generic-repeat-critical-threshold

jwchmodx commented Apr 3, 2026

Uh oh!

greptile-apps Bot commented Apr 3, 2026 •

edited

Loading

Comments Outside Diff (1)

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

vincentkoc commented Apr 28, 2026

Uh oh!

clawsweeper Bot commented Apr 28, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

jwchmodx commented Apr 3, 2026

Summary

Test plan

Uh oh!

greptile-apps Bot commented Apr 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Comments Outside Diff (1)

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

vincentkoc commented Apr 28, 2026

Uh oh!

clawsweeper Bot commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

greptile-apps Bot commented Apr 3, 2026 •

edited

Loading

clawsweeper Bot commented Apr 28, 2026 •

edited

Loading