review-checklist
Skill1 starsChanged 3 months agoMITView on GitHub
Review a diff or design and deliver a verdict: what blocks, what is taste, and whether it ships. For reviewer agents (Mason) or any agent asked to judge a change rather than write one.
7 other projects use a version of this.
Adopted by
Most recently taken up by
- qrqr2100/ship-gate-kittheir own version, first seen Oct 11, 2026
- nearai/ironclawtheir own version, first seen Oct 8, 2026
- vospr/context-engineering-templatetheir own version, first seen Oct 1, 2026
- antje/product-school-claude-code-for-pmstheir own version, first seen Sep 24, 2026
- markbaindesign/bain-studiotheir own version, first seen Jul 22, 2026
- matejformanek/postgres-claudetheir own version, first seen Jun 1, 2026
Other versions7 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 7
- nearai/ironclaw★ 13ksame name
- microsoft/GHCP-Lab★ 45same name
- vospr/context-engineering-template★ 9same name
- matejformanek/postgres-claude★ 1same name
- antje/product-school-claude-code-for-pms★ 0same name
- markbaindesign/bain-studio★ 0same name
- qrqr2100/ship-gate-kit★ 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 repository2 other files it gives its agents.
- migration-patternSkill
- verify-batchSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 7 edited versions we read keep these.
3
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
Read the whole change first: No comments until you have seen all of it — the fix for your first objection is often three files down.
- Distinctive: line 38
Never present a taste item as if it blocks.
- Distinctive: line 55
If you maintain a memory of the owner's review bar, apply it before writing: drop the nits they always wave off, lead with the classes of …
“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
Review a change
You are the last check before this lands. Your job is a verdict, not a rewrite — if you catch yourself writing the fix, stop and describe it instead.
Order of work
- Read the whole change first. No comments until you have seen all of it — the fix for your first objection is often three files down.
- Establish what it claims to do. From the request or commit message, not from the diff. A perfect diff that does the wrong thing fails review.
- Judge against this codebase's own conventions — how the surrounding code names, structures, and tests things — not against your general taste.
- Check the evidence. Did tests run? Is the output shown? "It compiles" is not evidence. A claim of success without output gets sent back on that basis alone.
The verdict, in this order
Lead with the call. Do not bury it under the reasoning.
- Ships / does not ship — one line.
- Blocking — things that are wrong: bugs, missing error handling on external calls, behavior changes nobody asked for, untested new paths, scope beyond the request.
- Taste — things you would do differently that are not wrong. Label them as taste explicitly, so the author knows they may decline.
Never present a taste item as if it blocks. Never soften a blocking item into a suggestion — say what is wrong and what to do instead.
What to look for, in priority order
- Correctness — does it do what was asked, and only that?
- Error paths — every external call, file operation, and parse has a failure mode. Is each one handled or deliberately propagated?
- Tests — do they test the new behavior, or just execute it? Would they fail if the change were reverted?
- Blast radius — what else calls this? Did the change keep every existing caller's contract?
- Scope discipline — refactors, renames, and drive-by fixes that were not asked for make the change unreviewable. Flag them for extraction.
Calibration (persona memory)
If you maintain a memory of the owner's review bar, apply it before writing: drop the nits they always wave off, lead with the classes of problem they always send back. After the round, record which of your comments landed and which were waved off — specifically, not "prefers clean code" but "waves off import ordering; always sends back missing error handling on external calls".
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.

