fix(tools): treat no-op writes and edits as terminal tool-loop failures (fixes #96983) by zw-xysk · Pull Request #97044 · openclaw/openclaw

zw-xysk · 2026-06-26T14:45:38Z

What Problem This Solves

When the write tool is called with content identical to the existing file, or the edit tool produces unchanged content, the agent loop treats these as progress and retries the same no-op mutation indefinitely.

Why This Change Was Made

write tool: readOriginalWriteState already detects identical content. Now returns a terminal no-op result with terminate: true instead of proceeding with the write.
edit tool: Filters out entries where oldText === newText before execution. If all entries are no-op, returns a terminal result. If only some are no-op, the valid sibling replacements still execute normally.
Also catches the No changes made error from edit-diff.ts and converts it to a terminal result.
Both tools use terminate: true to explicitly signal the tool-loop to stop retrying.

Changes Made

src/agents/sessions/tools/write.ts: Return {...textResult(...), terminate: true} when precheck detects identical content.
src/agents/sessions/tools/edit.ts: Filter out no-op entries (oldText === newText) before execution, preserving valid sibling edits. Catch No changes made error from edit-diff and return terminal result.
src/agents/sessions/tools/write.test.ts: Added 2 regression tests.
src/agents/sessions/tools/edit.test.ts: Added 3 regression tests (all no-op, mixed batch, real changes).

Evidence

Unit tests: node scripts/run-vitest.mjs run src/agents/sessions/tools/write.test.ts src/agents/sessions/tools/edit.test.ts — 14/14 passed

Real behavior proof

Environment: Linux, OpenClaw 2026.6.10, Node 24.13.1

WRITE IDENTICAL: No changes made to demo.txt. The file already has identical content. | terminate: true
WRITE DIFFERENT: Successfully wrote 12 bytes to demo.txt
EDIT NO-OP: No changes made to ... The replacement text is identical to the original. | terminate: true
EDIT MIXED BATCH: Successfully replaced 1 block(s) in ... (valid sibling preserved)
EDIT REAL CHANGE: Successfully replaced 1 block(s) in ...

✅ Write to existing file with identical content → no-op, terminate: true, file unchanged
✅ Write with different content → normal success
✅ Edit with all no-op entries → terminate: true, file unchanged
✅ Edit with mixed batch (no-op + real) → valid replacements still execute
✅ Edit with real changes → normal replace success

What was not tested: E2E agent loop with LLM calling write/edit (requires real model inference).

AI-assisted

zw-xysk · 2026-06-26T14:45:53Z

@clawsweeper re-review

clawsweeper · 2026-06-26T14:45:56Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

clawsweeper · 2026-06-26T14:46:48Z

Codex review: needs real behavior proof before merge. Reviewed June 26, 2026, 6:36 PM ET / 22:36 UTC.

Summary
The branch changes built-in write, edit, edit-diff, apply_patch, and focused tests so no-op file mutations can return terminal tool results instead of normal progress or retryable errors.

PR surface: Source +169, Tests +460. Total +629 across 7 files.

Reproducibility: yes. source-level reproduction is high confidence: current main allows the reported no-op write/edit/apply_patch behavior, and PR head would skip an exact trailing-whitespace edit because the fuzzy-normalized old/new strings compare equal.

Review metrics: 1 noteworthy metric.

Fuzzy normalization classes: 4 classes compared as no-op. Trailing whitespace, smart quotes, dashes, and Unicode spaces are normalized before the PR decides an edit is no-op, which can swallow real formatting edits.

Stored data model
Persistent data-model change detected: serialized state: src/agents/apply-patch.test.ts, serialized state: src/agents/apply-patch.ts, serialized state: src/agents/sessions/tools/edit-diff.ts, serialized state: src/agents/sessions/tools/edit.test.ts, serialized state: src/agents/sessions/tools/edit.ts, serialized state: src/agents/sessions/tools/write.test.ts, and 5 more. Confirm migration or upgrade compatibility proof before merge.

Root-cause cluster
Relationship: fixed_by_candidate
Canonical: #96983
Summary: This PR is an active candidate fix for the canonical no-op file-mutation issue; another open PR overlaps but is narrower, earlier attempts are closed, and the broad circuit-breaker issue only partially overlaps.

Members:

canonical: fix(tools): treat no-op edits and identical writes as terminal tool-loop failures #96983 - The PR body uses closing syntax for the exact issue requesting terminal handling for no-op write, edit, and patch-style mutations.
same_root_cause: Fix/no op file mutations #97008 - This open PR targets the same no-op write/edit failure mode but omits the latest apply_patch/fuzzy handling and is not established as the safer canonical landing path.
superseded: fix(tools): classify no-op file mutations #97000 - This earlier unmerged PR attempted the same no-op file-mutation classification and is now closed.
superseded: fix(tools): treat no-op writes and edits as terminal tool-loop failures (fixes #96983) #97004 - This earlier same-author PR was closed unmerged and replaced by this broader branch.
partial_overlap: Tool call circuit breaker needed — LLMs will blindly retry forever without one #78865 - The broad circuit-breaker issue shares the repeated tool-loop symptom but tracks retry, cooldown, and confirmation policy beyond this file-tool fix.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🧂 unranked krab
Proof: 🦪 silver shellfish
Patch quality: 🧂 unranked krab
Result: blocked until stronger real behavior proof is added.

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

Rank-up moves:

[P1] Fix fuzzy-equivalent no-op classification so exact formatting edits still apply.
[P1] Add redacted latest-head real behavior proof for apply_patch and the fuzzy edit edge case.
[P1] State or implement the intended structured no-op result contract for maintainers.

Proof guidance:

[P1] Needs stronger real behavior proof before merge: The PR body includes copied live output for write/edit, but latest-head apply_patch and fuzzy no-op behavior still need redacted terminal output, logs, a terminal screenshot, or another linked artifact; private details should be redacted before posting. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

[P1] The fuzzy no-op split can skip real exact edits that intentionally remove trailing whitespace or normalize Unicode punctuation/spacing, leaving the file unchanged while returning a terminal result.
[P1] Changing same-content write, no-change edit, and no-op apply_patch from success or ordinary errors into terminal results can alter upgraded session continuation behavior.
[P1] The PR body only shows copied write/edit output; latest-head apply_patch and fuzzy no-op behavior still need redacted real behavior proof.
[P1] The PR uses text plus terminate:true without stable structured no-op reason details, which may or may not be the contract maintainers want for UI, telemetry, or retry policy.

Maintainer options:

Fix fuzzy no-op handling before merge (recommended)
Keep exact formatting conversions as real edits, add a regression test for trailing-whitespace or Unicode normalization edits, and refresh latest-head proof for write/edit/apply_patch.
Require stable no-op details
Maintainers can require blocked/no-op reason details for write, edit, and apply_patch if downstream UI, telemetry, or retry policy needs a machine-readable contract.
Pause for the canonical issue path
If maintainers prefer a smaller or differently structured fix, pause this branch and keep the canonical issue as the landing target.

Next step before merge

[P1] Human review is needed because the PR has a blocking code regression plus contributor-owned proof and terminal-semantics gaps; no repair marker should be requested while real behavior proof is insufficient.

Security
Cleared: No concrete security or supply-chain concern was found; the diff changes built-in file-tool behavior and tests without dependency, workflow, credential, download, or package-execution changes.

Review findings

[P1] Preserve exact formatting edits before fuzzy no-op filtering — src/agents/sessions/tools/edit-diff.ts:466-467

Review details

Best possible solution:

Fix fuzzy no-op classification so exact formatting edits still apply, add latest-head proof for write/edit/apply_patch and the fuzzy edge case, then merge only after maintainers accept the terminal result semantics.

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

Yes, source-level reproduction is high confidence: current main allows the reported no-op write/edit/apply_patch behavior, and PR head would skip an exact trailing-whitespace edit because the fuzzy-normalized old/new strings compare equal.

Is this the best way to solve the issue?

No, not as currently written. The file-tool boundary is the right layer, but fuzzy-equivalent edits should only be terminalized after proving the actual applied content is unchanged or that oldText did not match exactly.

Full review comments:

[P1] Preserve exact formatting edits before fuzzy no-op filtering — src/agents/sessions/tools/edit-diff.ts:466-467
normalizeForFuzzyMatch strips trailing whitespace and normalizes quotes, dashes, and special spaces, so this branch puts exact edits like oldText: "foo " to newText: "foo" into noOpEdits. When no other real edit remains, edit.ts returns terminate:true without changing the file, even though current applyEditsToNormalizedContent would apply that exact replacement. Only classify fuzzy-equivalent pairs as no-op after confirming the exact old text was not present or that applying the edit leaves content unchanged.
Confidence: 0.92

Overall correctness: patch is incorrect
Overall confidence: 0.9

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against eba1ca683f6c.

Label changes

Label changes:

add rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🦪 silver shellfish and patch quality is 🧂 unranked krab.
remove rating: 🦪 silver shellfish: Current PR rating is rating: 🧂 unranked krab, so this older rating label is no longer current.

Label justifications:

P2: The PR addresses a normal-priority agent file-tool loop bug, but the current patch has bounded correctness and proof blockers before merge.
merge-risk: 🚨 compatibility: The PR changes existing file-tool outcomes from success or ordinary errors to terminal results and can skip formatting edits that current behavior applies.
merge-risk: 🚨 session-state: The patch changes whether file-tool calls stop the current tool batch, which affects agent session continuation and false-progress handling.
rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🦪 silver shellfish and patch quality is 🧂 unranked krab.
status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs stronger real behavior proof before merge: The PR body includes copied live output for write/edit, but latest-head apply_patch and fuzzy no-op behavior still need redacted terminal output, logs, a terminal screenshot, or another linked artifact; private details should be redacted before posting. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence reviewed

PR surface:

Source +169, Tests +460. Total +629 across 7 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	4	202	33	+169
Tests	3	465	5	+460
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	7	667	38	+629

What I checked:

Repository policy applied: Root AGENTS.md and scoped src/agents/AGENTS.md were read; the review-depth, session-state, compatibility, and proof guidance applies because this PR changes agent file-tool loop behavior. (AGENTS.md:11, eba1ca683f6c)
Current main write gap: Current main can detect same-content writes with readOriginalWriteState, but still calls writeFile and reports successful progress. (src/agents/sessions/tools/write.ts:395, eba1ca683f6c)
Current main edit behavior: Current main's edit path applies exact replacements and throws a no-change error only after baseContent equals newContent, so exact oldText/newText differences such as trailing-whitespace removal are real edits before this PR's fuzzy prefilter. (src/agents/sessions/tools/edit-diff.ts:199, eba1ca683f6c)
Terminal result contract: AgentToolResult.terminate is a batch-stop hint, and the loop stops only when every finalized tool result has terminate:true, so changing file tools to terminal results is session-behavior-sensitive. (packages/agent-core/src/types.ts:449, eba1ca683f6c)
PR head fuzzy no-op regression: PR head classifies edits as no-op when normalizeForFuzzyMatch(oldText) equals normalizeForFuzzyMatch(newText), which includes trailing whitespace, smart quotes, dashes, and special spaces rather than only exact no-ops. (src/agents/sessions/tools/edit-diff.ts:466, be67a514a2a6)
PR head terminalizes filtered edits: After splitNoOpEdits, edit.ts returns terminate:true when realEdits is empty, so an exact edit like removing trailing spaces can be reported as no changes and skipped. (src/agents/sessions/tools/edit.ts:430, be67a514a2a6)

Likely related people:

kevinslin: The current write/edit/apply_patch files in this checkout trace through commit a6a4652 and merged PR fix(codex): expose plugin apps after delayed inventory load #96872, authored by kevinslin. (role: introduced current tool implementation; confidence: high; commits: a6a4652c708a; files: src/agents/sessions/tools/edit-diff.ts, src/agents/sessions/tools/edit.ts, src/agents/sessions/tools/write.ts)
vincentkoc: Vincent force-pushed the latest head and authored the later commits that added apply_patch and fuzzy edit no-op handling on this branch. (role: recent follow-up owner; confidence: high; commits: 6c20b616d471, 63000f6aaf4d, be67a514a2a6; files: src/agents/apply-patch.ts, src/agents/sessions/tools/edit-diff.ts, src/agents/sessions/tools/edit.ts)
steipete: Live PR timeline shows this PR was assigned to steipete during the maintainer review window; this is routing context rather than blame for the code. (role: assigned reviewer; confidence: medium; files: src/agents/sessions/tools/edit-diff.ts, src/agents/apply-patch.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.

zw-xysk · 2026-06-26T14:56:27Z

@clawsweeper re-review

clawsweeper · 2026-06-26T14:56:31Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

zw-xysk · 2026-06-26T15:13:18Z

@clawsweeper re-review

clawsweeper · 2026-06-26T15:13:21Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

vincentkoc · 2026-06-26T22:41:31Z

Merged via squash.

Prepared head SHA: be67a514a2a6b39206fceb0b57335cb9caacd313
Landed commit: ab966c214be08d99c23695f496c6688bcc12486a

…es (fixes openclaw#96983) (openclaw#97044) * fix(tools): treat no-op writes and edits as terminal tool-loop failures Fixes openclaw#96983 * fix(tools): treat no-op writes and edits as terminal tool-loop failures Fixes openclaw#96983 * fix(tools): preserve valid sibling edits in mixed no-op batches Fixes openclaw#96983 * fix(tools): terminate apply_patch no-ops safely * fix(tools): validate no-op edits independently * fix(tools): preserve no-op edit overlap checks * fix(tools): preserve no-op patch file formatting * fix(tools): preserve move no-op formatting * fix(tools): narrow same-path move no-op typing * fix(tools): distinguish edit no-op errors * fix(tools): keep edit previews aligned with execution * fix(tools): align no-op validation and formatting * fix(tools): preserve empty patch and preview no-ops * fix(tools): preview fuzzy edit no-ops cleanly * fix(tools): isolate fuzzy-equivalent edit no-ops --------- Co-authored-by: Vincent Koc <[email protected]>

…es (fixes openclaw#96983) (openclaw#97044) * fix(tools): treat no-op writes and edits as terminal tool-loop failures Fixes openclaw#96983 * fix(tools): treat no-op writes and edits as terminal tool-loop failures Fixes openclaw#96983 * fix(tools): preserve valid sibling edits in mixed no-op batches Fixes openclaw#96983 * fix(tools): terminate apply_patch no-ops safely * fix(tools): validate no-op edits independently * fix(tools): preserve no-op edit overlap checks * fix(tools): preserve no-op patch file formatting * fix(tools): preserve move no-op formatting * fix(tools): narrow same-path move no-op typing * fix(tools): distinguish edit no-op errors * fix(tools): keep edit previews aligned with execution * fix(tools): align no-op validation and formatting * fix(tools): preserve empty patch and preview no-ops * fix(tools): preview fuzzy edit no-ops cleanly * fix(tools): isolate fuzzy-equivalent edit no-ops --------- Co-authored-by: Vincent Koc <[email protected]> (cherry picked from commit ab966c2)

openclaw-barnacle Bot added agents Agent runtime and tooling size: S labels Jun 26, 2026

vincentkoc assigned steipete Jun 26, 2026

openclaw-barnacle Bot added size: M and removed size: S labels Jun 26, 2026

openclaw-barnacle Bot added size: L and removed size: M labels Jun 26, 2026

clawsweeper Bot removed the rating: 🦐 gold shrimp Decent PR readiness signal, but merge confidence is limited. label Jun 26, 2026

vincentkoc added 11 commits June 26, 2026 15:24

fix(tools): terminate apply_patch no-ops safely

6c20b61

fix(tools): validate no-op edits independently

37e4581

fix(tools): preserve no-op edit overlap checks

2f72771

fix(tools): preserve no-op patch file formatting

2d35c14

fix(tools): preserve move no-op formatting

f798b7e

fix(tools): narrow same-path move no-op typing

2d17e2d

fix(tools): distinguish edit no-op errors

7f89c8b

fix(tools): keep edit previews aligned with execution

6a219c5

fix(tools): align no-op validation and formatting

ffe7a4b

fix(tools): preserve empty patch and preview no-ops

a4bd129

fix(tools): preview fuzzy edit no-ops cleanly

63000f6

vincentkoc force-pushed the fix/issue-96983-noop-tools-v2 branch from c2b4903 to 63000f6 Compare June 26, 2026 22:24

fix(tools): isolate fuzzy-equivalent edit no-ops

be67a51

clawsweeper Bot added rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. and removed rating: 🦪 silver shellfish Thin PR readiness signal; proof, validation, or implementation needs work. labels Jun 26, 2026

vincentkoc merged commit ab966c2 into openclaw:main Jun 26, 2026
105 of 107 checks passed

github-actions Bot mentioned this pull request Jun 26, 2026

📡 Upstream Digest — 2026-06-26 22:59 UTC curtismercier/openclaw-mods#1156

Open

clawsweeper Bot mentioned this pull request Jun 27, 2026

fix(edit): unwrap XML-RPC {item: {oldText, newText}} in prepareArguments before validation #96700

Closed

This was referenced Jun 28, 2026

improve: show edit mismatch candidate lines #97512

Closed

feat(edit): show closest candidate lines on oldText mismatch (#97032) #97305

Closed

fix(agents): classify already-at-target edit follower as noop (#60816) #77638

Open

This was referenced Jul 1, 2026

fix(edit): preserve unrelated lines during fuzzy text matching #90060

Closed

fix(edit): keep mismatch hint truncation unicode-safe #99527

Merged

improve: show edit mismatch candidate lines #97269

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(tools): treat no-op writes and edits as terminal tool-loop failures (fixes #96983)#97044

fix(tools): treat no-op writes and edits as terminal tool-loop failures (fixes #96983)#97044
vincentkoc merged 15 commits into
openclaw:mainfrom
zw-xysk:fix/issue-96983-noop-tools-v2

zw-xysk commented Jun 26, 2026 •

edited

Loading

Uh oh!

zw-xysk commented Jun 26, 2026

Uh oh!

clawsweeper Bot commented Jun 26, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 26, 2026 •

edited

Loading

Uh oh!

zw-xysk commented Jun 26, 2026

Uh oh!

clawsweeper Bot commented Jun 26, 2026 •

edited

Loading

Uh oh!

zw-xysk commented Jun 26, 2026

Uh oh!

clawsweeper Bot commented Jun 26, 2026 •

edited

Loading

Uh oh!

Uh oh!

vincentkoc commented Jun 26, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Uh oh!

Conversation

zw-xysk commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What Problem This Solves

Why This Change Was Made

Changes Made

Evidence

Real behavior proof

Uh oh!

zw-xysk commented Jun 26, 2026

Uh oh!

clawsweeper Bot commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

zw-xysk commented Jun 26, 2026

Uh oh!

clawsweeper Bot commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

zw-xysk commented Jun 26, 2026

Uh oh!

clawsweeper Bot commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

vincentkoc commented Jun 26, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

zw-xysk commented Jun 26, 2026 •

edited

Loading

clawsweeper Bot commented Jun 26, 2026 •

edited

Loading

clawsweeper Bot commented Jun 26, 2026 •

edited

Loading

clawsweeper Bot commented Jun 26, 2026 •

edited

Loading

clawsweeper Bot commented Jun 26, 2026 •

edited

Loading