review-fix
Skill60 starsChanged 34 days agoMITView on GitHub
Apply explicitly accepted, evidence-backed defect fixes in the eligible existing session that produced the review or diagnosis, preserving secured evidence.
11 other projects use a version of this.
Adopted by
Most recently taken up by
- nwiizo/ccswarmtheir own version, first seen Nov 6, 2026
- davidvornholt/mail-mcptheir own version, first seen Nov 5, 2026
- davidvornholt/mailboxtheir own version, first seen Nov 5, 2026
- raine/skillstheir own version, first seen Nov 1, 2026
- 200iqlabs/shared-skillstheir own version, first seen Oct 31, 2026
- pcode-xyz/simple-skillstheir own version, first seen Oct 20, 2026
Used with
Projects with it also use review.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
SentrySkillreview
Fixture review skill.
Also in 5 projects that use this
Other versions11 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 11
- nwiizo/ccswarm★ 156same name
- enes-alatas/spool★ 48same name
- jhd3197/CachiBot★ 19same name
- EternallLight/tgc-skills★ 11same name
- 200iqlabs/shared-skills★ 4same name
- pcode-xyz/simple-skills★ 4same name
- davidvornholt/mail-mcp★ 2same name
- davidvornholt/mailbox★ 2same name
- raine/skills★ 2same name
- marcuskrogh/skills★ 1same name
1 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository17 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/agents.mdAGENTS.md
- archive-legacy-docsSkill
- author-and-verify-implementation-planSkill
- browser-walkthroughSkill
- create-planSkill
- delegate-analysisSkill
- execute-work-packageSkill
- generate-docsSkill
- generate-handoverSkill
- report-environment-issueSkill
- resume-planSkill
- review-implementation-planSkill
- review-implementationSkill
- review-planSkill
- update-docsSkill
- update-planSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. 0 are common, 12 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 10 edited versions we read keep these.
12
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 27
Do not use for:
- Very specific: line 55
Use tpl-review-fix-prompt.md for an eligible continuation.
- Very specific: line 59
If the source session's retained review/diagnosis context/output is correct and sufficient, resume its exact task_id with the same persona …
- Distinctive: line 66
Read the finding source and reuse the secured evidence in the existing session context.
- Distinctive: line 85
Do not modify a source review artifact.
- Distinctive: line 86
Do not modify phase scope, acceptance criteria, or unrelated docs/plans.
- Distinctive: line 87
Do not perform Git history operations.
- Distinctive: line 97
Next: stop | primary decision required.
- Distinctive: line 111
Follow the session eligibility and fallback rules above.
- Distinctive: line 112
Keep the same persona and model/variant while resuming.
- Distinctive: line 113
Do not re-open already rejected findings unless the primary explicitly asks for reconsideration.
- Distinctive: line 114
Do not describe self-verification as an independent review.
- Distinctive: line 115
Never weaken, delete, or skip tests to make verification pass.
- Distinctive: line 116
Do not replace removed gold-plating with new abstraction, infrastructure, tests, or documentation unless an accepted finding and the …
“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skill: Review Fix
This skill provides context-preserving defect remediation: apply explicitly accepted, evidence-backed fixes using the eligible existing session that produced the review or diagnosis. A review artifact is one supported finding source, not a prerequisite.
The point of preserving context is to reuse already-secured evidence, not collect it again. Related fixes may span files, call sites, and tests within the accepted objective. Reuse depends on correct and sufficient retained context/output, not the source workflow's mode or fix size.
When to Use
Use after:
review-implementation-planidentifies corrections to an implementation plan.review-implementationidentifies code, test, configuration, or integration corrections.- An analysis digest/message identifies a concrete defect and its diagnosis.
- An Implementer digest/message identifies a concrete defect of the same completed work package.
Do not use for:
- Initial review work.
- Plan review corrections; plans remain conversation-owned by the primary. Resume the reviewer for clarification, then update the plan through
update-plan. - A genuinely new objective, changed gated scope, new dependency decision, or user-requested independent perspective.
| Situation | Route |
| Accepted defect, explicit Primary remediation instruction, eligible source session | review-fix in that existing session |
| New implementation objective/package, or fresh implementation needed under the fallback below | execute-work-package in a fresh Implementer |
| New technical implementation-plan work, or fresh technical planning needed under the fallback | author-and-verify-implementation-plan in a fresh session |
| Conversation-owned plan corrections, including phase scope or acceptance changes | Primary via update-plan; reviewer may clarify, not edit the plan |
| Mechanical self-fix within the Maintainer's existing narrow allowance | Maintainer self-fix; no expansion of that allowance |
| Evidence retrieval | retriever may supply evidence, but is not a remediation owner |
Required Inputs
The continuation prompt must provide:
- Finding source: review artifact path or analysis/implementation digest or message reference; multiple sources may be bundled.
- Accepted defect set: IDs when present, otherwise stable short labels; include each defect's concrete description, expected behavior/invariant, and evidence.
- Authority: the Primary has accepted the defects and explicitly instructed remediation within the supplied scope. This instruction is sufficient; no approval token or new Blueprint is required.
- Session: exact source
task_id, same persona and model/variant, and correct and sufficient retained review/diagnosis context/output for the assigned defects. - Target type:
implementation-planorimplementation. - Scope: accepted objective, relevant boundaries, and Primary decisions. Do not require an exact file allowlist when related call sites or tests need to be discovered.
- Verification: command or expected verification goal. Identify whether a supplied command is broad/full. The fixer may choose focused targeted tests when none were supplied.
- Closure: one bounded remediation pass, compact digest, then stop and return unresolved defects or decisions to the Primary.
Use tpl-review-fix-prompt.md for an eligible continuation.
Session eligibility and fallback
- If the source session's retained review/diagnosis context/output is correct and sufficient, resume its exact
task_idwith the same persona and model/variant throughreview-fix. - If that context/output is faulty, insufficient, or unavailable, the Primary starts a fresh session through the appropriate route above, with a pointer to the finding source as a hint. The fresh session must establish the evidence it needs; never pretend it has the old context. This is a fresh workflow, not
review-fixsession reuse. - An Implementer may be resumed after its digest only for accepted defects of the same completed package under this contract. This narrow retirement exception cannot start another phase/package or authorize indefinite continuation.
- The Primary may bundle findings from multiple source sessions in one pass and assigns defects to the eligible sessions explicitly. Each resumption retains its own exact
task_idand persona/model variant; pointers to other sources do not confer their retained context. Return each assigned digest to the Primary; bundling does not create an automatic continuation loop.
Protocol
- Read the finding source and reuse the secured evidence in the existing session context. Inspect current state or evidence gaps only as needed for the fix; do not repeat broad evidence collection.
- Confirm explicit Primary authorization, session eligibility, and that the fixes still serve the accepted objective. If eligibility fails, stop and report the fallback need to the Primary before dependent edits.
- Apply the necessary related corrections. For implementation-plan scope findings, remove, merge, or simplify the accepted steps/artifacts instead of adding compensating design. Size alone does not require a new work package.
- For implementation fixes, use targeted tests while editing and preserve any supplied broad/full final gate. For implementation-plan fixes, check only the changed steps and directly affected references; do not invent a formal verification suite.
- Return the compact digest below and stop. This is one remediation pass, not the start of a review loop.
Remediation posture
No Gold-Plating. No Adversarial Reviewing. No Scope Creep. Fix accepted, evidence-backed findings and the related changes required to complete them. Do not use remediation to invent improvements, reopen rejected findings, or create another review/fix cycle.
The Primary's explicit instruction naming the accepted defects and scope is the gate; acceptance alone is not an instruction to edit. Do not add unrelated improvements or silently change the objective. If a new decision is required, report it to the Primary; do not invent a new gate or switch agents autonomously.
Write Boundary
- For
implementation-plan, edit the relevant implementation-plan artifacts and related references required by the accepted findings. - For
implementation, edit related code, configuration, integration points, and tests required by the accepted findings. - Do not modify a source review artifact; it remains the immutable record of the independent review.
- Do not modify phase scope, acceptance criteria, or unrelated docs/plans.
- Do not perform Git history operations.
Output Contract
Return only:
- Outcome: succeeded | partial | blocked
- Findings: fixed and unresolved defect IDs or stable short labels
- Edits: changed files with one-line descriptions
- Verify: command/checks and result
- Next: stop | primary decision required; an independent re-review was not started
Independent Re-review
The remediation pass is not an independent review. A fresh review is optional and requires an explicit primary or user decision. The primary may choose it when:
- the primary wants an independent final opinion;
- the fix exposes uncertainty that the current fixer cannot resolve;
- security, persistence, public API, or migration risk warrants another perspective.
Do not start or recommend a fresh review merely because findings were fixed, steps were removed, or the change spans several files. Do not chain review -> fix -> review -> fix. After one remediation pass, stop. Materially new risk/uncertainty must be escalated to the Primary, who decides whether to request an independent re-review; it never triggers one automatically.
Rules
- Follow the session eligibility and fallback rules above; never substitute a fresh session while claiming retained context.
- Keep the same persona and model/variant while resuming.
- Do not re-open already rejected findings unless the primary explicitly asks for reconsideration.
- Do not describe self-verification as an independent review.
- Never weaken, delete, or skip tests to make verification pass.
- Do not replace removed gold-plating with new abstraction, infrastructure, tests, or documentation unless an accepted finding and the accepted objective require it.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

