fix-findings
Skill20 starsChanged 52 days agoMITView on GitHub
Apply every finding from the most recent review, audit, verification, or check in this session.
8 other projects use a version of this.
Adopted by
Most recently taken up by
- snuffish/dotfilestheir own version, first seen Nov 8, 2026
- chrislacey89/skillstheir own version, first seen Nov 4, 2026
- mondoohq/secure-stacktheir own version, first seen Nov 1, 2026
- wynn-dev/skillstheir own version, first seen Oct 20, 2026
- abdurshd/skillstheir own version, first seen Oct 13, 2026
- quaat/autonomous-developmenttheir own version, first seen Sep 24, 2026
Used with
Projects with it also use codex-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.
juriwiensSkillcodex-review
Run code review via Codex CLI
Also in 3 projects that use this
Other versions8 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 8
- chrislacey89/skills★ 4same name
- quaat/autonomous-development★ 3same name
- SemanticMatter/autonomous-development★ 3same name
- abdurshd/skills★ 2same name
- mondoohq/secure-stack★ 2same name
- renatodvc/rstack★ 1same name
- wynn-dev/skills★ 1same name
- snuffish/dotfiles★ 0same name
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 repository41 other files it gives its agents.
- AGENTS.mdAGENTS.md
- ask-supermattSkill
- babysit-prSkill
- codebase-designSkill
- code-reviewSkill
- commitSkill
- dependency-vettingSkill
- diagnosing-bugsSkill
- domain-modelingSkill
- grillingSkill
- grill-meSkill
- grill-with-docsSkill
- handoffSkill
- ideateSkill
- implementSkill
- implement-specSkill
- improve-architectureSkill
- improve-file-structureSkill
- improve-testsSkill
- jurySkill
- orchestrateSkill
- prototypeSkill
- prSkill
- qaSkill
- recover-ciSkill
- refactorSkill
- researchSkill
- resolving-merge-conflictsSkill
- retroSkill
- setup-supermatt-skillsSkill
- ship-prSkill
- tddSkill
- teachSkill
- to-questionnaireSkill
- to-specSkill
- to-ticketsSkill
- triageSkill
- wait-whatSkill
- wayfinderSkill
- what-would-you-doSkill
- writing-for-agentsSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 7 edited versions we read keep these.
1
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 11
No such report in this session and no finding text in the arguments: say so and stop.
- Very specific: line 12
Before the first edit in each affected repository, discover and read its applicable AGENTS.md, CLAUDE.md, and CONTRIBUTING.md, including …
- Very specific: line 18
Do not commit. The user commits with /commit or /ship-pr in Claude linked installs or Cursor, /supermatt:commit or /supermatt:ship-pr from …
“Common” is measured: kept by at least 6 in 10 of the 7 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
Apply the findings from the most recent review, audit, verification, lint run, test run, or other check reported earlier in this session, whatever produced it. Arguments that name findings (codes or numbers) narrow which ones; arguments that carry a finding's own text are the report, even when it came from another session. No arguments means every finding, at every severity. A finding the report left as a choice for the user: apply the option the arguments name, else the one the report recommended in so many words ("I recommend X"), worded exactly as it was offered. An option listed first or labelled "proposed" is not a recommendation. With none named, apply no option: ask the caller which one to take, and report the finding as <code>: pending, <the choice it waits on> until the caller answers. The out-of-scope deferral in step 4 is the one open choice with its own form, <code>: deferred, follow-up pending.
- No such report in this session and no finding text in the arguments: say so and stop. Do not go looking for problems to invent.
- Before the first edit in each affected repository, discover and read its applicable
AGENTS.md,CLAUDE.md, andCONTRIBUTING.md, including any instructions they point to. Repeat this before touching another repository. Record absent instruction files separately from a repository whose available instructions document no checks. - Fix root causes, not the symptom each finding names. Grep every caller before editing a shared function: one guard where the callers converge beats a guard in each of them.
- Do not expand scope. A finding is a fix, not an invitation to refactor around it. If its fix needs a schema or data migration, a new ADR, or a redesign beyond the branch's change (a new subsystem, scheduler, or cross-module restructuring), check whether the defect already exists at the merge base with the default branch. If it does, the fix is out of scope for this branch: do not apply it. Offer filing a follow-up issue and say "I recommend a follow-up issue". Until the caller chooses, report the finding as
<code>: deferred, follow-up pending(this form, notpending, even when the report also left the finding as a choice), and name no issue. When the follow-up is chosen, create it with the Create an issue operation indocs/agents/issue-tracker.md, orgh issue createwhen that file is missing, including the defect, its evidence, and the proposed fix, then report the finding as<code>: deferred to <issue>. When it is declined, report<code>: deferred, follow-up declined. A defect introduced by this branch is still fixed on the branch, whatever the size, following the rules below. For in-scope findings, when the finding proposes a fix, apply that fix, together with any change it cannot work without, and name that change on the finding's report line. If it cannot work even so, skip the finding and say why instead of choosing another. If it works at only some of the places the finding names, apply it there and report the finding as partly fixed. - For every edited repository, run its required checks and make its other required edits (a version bump, a changelog entry). Done when each edited repository has its checks run, or the report distinguishes unavailable instructions from available instructions that document no checks.
- Report one line per finding:
<code>: fixed,<code>: partly fixed, <what is left and why>,<code>: pending, <the choice it waits on>,<code>: deferred, follow-up pending,<code>: deferred to <issue>,<code>: deferred, follow-up declined, or<code>: skipped, <why>. Then one line per edited repository with the checks run. Say plainly when a finding was wrong.
Do not commit. The user commits with /commit or /ship-pr in Claude linked installs or Cursor, /supermatt:commit or /supermatt:ship-pr from the Claude plugin, $commit or $ship-pr in Codex, or /skill:commit or /skill:ship-pr in OMP.
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.

