ship-check
Skill2 starsChanged 41 days agoMITView on GitHub
Pre-ship checklist for any script, tool, or repo the agent writes or packages, run before code leaves the session and before the handback is written: path traversal, command injection, secrets and names, placeholders, .gitignore, license, atomic writes, and proof that every guard has been seen to fail. Use it whenever a session is about to hand back a change that ships executable code, and name the sections that did not apply.
19 other projects use a version of this.
Adopted by
Most recently taken up by
- Archive228/claude-fable-rea-blatheir own version, first seen Nov 15, 2026
- nadavnaveh/skillstheir own version, first seen Nov 8, 2026
- kitloomlk-stack/nextjs-supabase-kittheir own version, first seen Oct 15, 2026
- textura-agency/next16-claude-startertheir own version, first seen Oct 14, 2026
- joetawil7/first-passtheir own version, first seen Oct 14, 2026
- AhmedMerza/agent-skillstheir own version, first seen Oct 11, 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 3 projects that use this
Other versions19 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 19
- facontidavide/PlotJuggler★ 6.3ksame name
- textura-agency/next16-claude-starter★ 141same name
- joetawil7/first-pass★ 128same name
- Archive228/claude-fable-rea-bla★ 18same name
- mohankrishnaalavala/context-router★ 11same name
- Aximande/agent-skills★ 4same name
- koach08/claude-code-skills★ 4same name
- nadavnaveh/skills★ 3same name
- AhmedMerza/agent-skills★ 2same name
- AhmedMerza/claude-skills★ 2same name
9 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 repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- vault-relaySkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 18 edited versions we read keep these.
2
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
- Distinctive: line 8
Run this before any script, tool, or repo you wrote leaves your hands, and before the handback is written.
- Very specific: line 21
NOT IN THE FILES. Secrets, referenced by location and rotation path, never by value.
- Distinctive: line 51
BEFORE PACKAGING. Run sections 1, 2 and 5 on the final tree, not the draft.
“Common” is measured: kept by at least 6 in 10 of the 18 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
Ship check
Run this before any script, tool, or repo you wrote leaves your hands, and before the
handback is written. State the result in the handback, never as a bare "clean": what was
fixed, or "clean" followed by which sections did not apply and why. A check run silently is
indistinguishable from one skipped. The reasoning behind each item is in SHIP-CHECK.md at
the root of the claude-code-discipline package
(https://github.com/fightingforsidewalk/claude-code-discipline).
- INPUT THAT BECOMES SOMETHING ELSE. Paths: allow-list the characters AND resolve-and-contain
the final path, tested with
../../x,x/../../y,... Shell: argument lists, never a string built from input. Format strings, regex, SQL, HTML: ask of every input what language it lands in, and escape or parameterise for that language. File content that is also syntax (comment delimiters, fences, front-matter): refuse or escape. - NOT IN THE FILES. Secrets, referenced by location and rotation path, never by value; grep
for
key,token,secret,password,://. Names, when the artefact is for others: people, companies, products, internal roles, local paths. Unresolved placeholders (TODO,TBD,<your-handle>,example.com): fill, remove, or name as the human's to fill. Session-specific paths. Invisible characters: scan the final bytes for anything below0x20that is not newline or tab, plus0x7F, U+2028 and U+2029, and write them in source as escape sequences rather than literals. - REPO HYGIENE.
.gitignorepresent and not ignoring folders the human's data will live in.LICENSEpresent and its holder intentional. Every referenced file exists, README links resolve, the layout section matches the tree's visible files — dotfiles and repo housekeeping (.gitignore,.editorconfig,.github/and the like) are not listed there — a section number cited in one file exists in the other. Front-matter parses: a Markdown file with YAML front-matter goes through a parser before packaging, never an eyeball, and adescription:value carrying a colon-space is quoted, or YAML reads the colon as a nested key and a strict loader refuses the file. Help text and docstrings match current behaviour. - WRITES AND FAILURE. Atomic writes: temp file then replace. Assert invariants before mutating, so a failed assertion leaves the file untouched. Fail loudly with a reason a human can act on. Refuse the dangerous case rather than warn, when refusing is free.
- PROVE THE GUARDS, IN ORDER. Feed every refusal the input it should refuse and show the refusal in the handback; then run the happy path once more. When FIXING a defect: reproduce it against the unfixed code and show it, then fix, then show it refused. Verifying only after the fix cannot distinguish "the guard works" from "the defect was never there". If reproduction is impractical, say so and say what was checked instead. Applies to correctness and security claims, not to docs, copy or renames. A comment asserting a verification is itself a claim and carries who established it and how.
- BEFORE PACKAGING. Run sections 1, 2 and 5 on the final tree, not the draft. List every file going out. Name any section that did not apply and why. If a second reviewer exists (another chat, a linter, a scanner), say so and invite 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.

