correct
Skill1 starsChanged 33 days agoMITView on GitHub
Turn a repeated mistake into a check that fails on it. When the owner makes the same correction twice, or a reviewer or verifier finds the same class of problem twice, add the highest check that catches it (a type, then a lint whose message names the fix, then a test, then a verifier rule, then a doc), prove it fails on the past mistake, and keep the repo's AGENTS.md rule → check table. Use whenever the owner corrects you, when a finding repeats, and for 'never again' or 'make this impossible'.
50 other projects use a version of this.
Adopted by
Most recently taken up by
- Yeachan-Heo/oh-my-claude-sisyphustheir own version, first seen Nov 8, 2026
- Loner1024/skillstheir own version, first seen Nov 8, 2026
- maxatwork/skillstheir own version, first seen Nov 8, 2026
- caidanw/skillstheir own version, first seen Nov 8, 2026
- painhardcore/pstacktheir own version, first seen Nov 8, 2026
- Yeachan-Heo/oh-my-claudecodetheir own version, first seen Nov 8, 2026
Used with
Projects with it also use ketchup-video, poteto-help and 10 more.
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.
tristanphvnSkillketchup-video
Catch the user up like /keel:ketchup, with a short silent video of slides for what happened and the waiting calls as text under it. Use for /keel:ketchup-video.
Also in 13 projects that use this
ianclearySkillpoteto-help
Guides users through pstack setup, /poteto-mode, and picking the skill, playbook, or principle for a task. Type /poteto-help with a question.
Also in 14 projects that use this
ianclearySkillprinciple-explain-the-number
Apply before you trust, report, or act on a number you measured: a speedup, a regression, a throughput, a latency, or an eval result. Find what limits it, and rule out that it measured something other than the work you think.
Also in 27 projects that use this
Xia-AtaraxiaSkillbenchmark-checklist
Vets performance measurements for the limiter, tuning, physical limits, errors, repeatability, relevance, and proof that timed work happened. It applies to 'check this benchmark', 'report this speedup', measured regressions, and choosing a library or configuration from numbers. It reports spread and gaps before drawing a conclusion. Not for general testing policy or choosing test oracles - use principle-testing.
Also in 36 projects that use this
instructureSkillketchup
Rebuild context after /clear by analyzing uncommitted changes across Canvas LMS and all plugin repositories
Also in 23 projects that use this
drafaelSkillin-your-own-words
Restate the user's goals and the problem they are trying to solve to check understanding. Manual invocation only; use when the user explicitly invokes in-your-own-words or requests this understanding check. Do not solve the task or rewrite the assistant's last answer.
Also in 23 projects that use this
Other versions50 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 50
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- backnotprop/pstack★ 3.8ksame name
- moeru-ai/xsai★ 667same name
- savannahfeder/agentbox★ 376same name
- jtaroreh/agystack★ 128same name
- uzairansaruzi/p3-stack★ 126same name
- dsebban/skills★ 71same name
- TheOnlyFusionCube/potetos-for-everyone★ 21same name
- creedants/pstack-t3★ 18same name
40 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 repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- blast-radiusSkill
- claude-remote-control-serverSkill
- councilSkill
- cross-reviewSkill
- docs-reviewSkill
- evidence-bundleSkill
- herdrSkill
- intent-cardSkill
- one-termSkill
- owner-charterSkill
- quotaSkill
- starbridgeSkill
- test-auditSkill
- unnarrateSkill
- unslopSkill
- verifySkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.
0
Common
Most of the 16 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
First time: add a row with Check: none yet and the PR or message that showed it.
- Distinctive: line 42
Run the check on the real past mistake: check out the bad commit in a scratch worktree, or reapply its diff, and paste the failing output.
“Common” is measured: kept by at least 6 in 10 of the 16 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
Correct
Assume the next contributor is an agent that sees only the files it opened, copies the nearest example, and takes the shortest path that compiles. A rule it has to remember fails; a check that fails its build doesn't.
On every correction
- Fix the instance.
- Look it up in the repo's rule table (AGENTS.md, "Rules and their
checks"). Then search for earlier cases:
git log -S/--grep, review threads (gh api repos/<o>/<r>/pulls/comments), pastcross-reviewandverifyfindings. - First time: add a row with
Check: none yetand the PR or message that showed it. The row is how the second time gets noticed. - Second time (a row already there, or an earlier case found): build the check now, in this change or the next PR.
Pick the highest check that catches it
- Type: make the bad state impossible to write. When the mistake comes from two ways of doing one thing, delete one; then there is nothing to get wrong.
- Lint or CI check whose message names the fix: "use
formatTime()fromshared/time.ts, nottoLocaleTimeString", not "invalid time format". If the pattern is already common, fail only on lines a change adds. - Test of the behaviour, through the
test-auditgate. - Verifier rule: for what only a render or a run shows (off-centre,
clipped, wrong in one locale), a rule in the repo's
verify-<app>skill that says what to measure and the tolerance. - Doc: an AGENTS.md line, only for a judgment call no check can make. Nothing fails when an agent skips it.
Prove it
Run the check on the real past mistake: check out the bad commit in a scratch worktree, or reapply its diff, and paste the failing output. Then run it on the fix and show it pass. The same command runs locally and in CI.
The table
## Rules and their checks
| Rule | Check | Seen in |
|---|---|---|
| App icon centred against its peers | verify-android rule "icon centre" | #41, #57 |
| UI strings carry no AI tells | `pnpm lint:copy` | #63 |
| A Settings subtitle only when the title can't stand alone | none yet | #70 |A row whose check is "none yet" and that is seen again is a repeat: build the check in the same change. Drop a row once its mistake can't happen.
Gotchas
- A message that names no fix gets the check suppressed or worked around by the next agent.
- Proving on an example you made up misses what the real mistake looked like (a regex that catches your example and not the past diff).
- A doc-level fix when a check was possible is how the same correction comes a third time.
- Exceptions go on the offending line, with the reason.
- A broad exemption (a whole directory) turns the check off. Fail only on new lines instead.
- One check per commit, naming the correction it encodes, so a revert is clean.
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.

