ship-a-feature
Skill15 starsChanged 5 months agoMITView on GitHub
Ship a NEW capability to vigiles so it actually reaches a user — a new export, public function, subpath, CLI flag, or hook-vocabulary word. Use when about to add, or having just added, a feature / export / public API to this repo. Runs the executable checks a green `npm run check` does NOT make — reachable from a public door, API-surface diff READ, a public doc home, the experimental_ decision, a gate on a frozen tree. NOT for bug fixes, refactors, doc-only or test-only edits.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- VodouAI/OStheir own version, first seen Jul 19, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- VodouAI/OS★ 1same 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 repository24 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- examples/CLAUDE.mdCLAUDE.md
- src/CLAUDE.mdCLAUDE.md
- add-a-linterSkill
- audience-checkSkill
- audit-feedback-loopSkill
- code-qualitySkill
- deep-researchSkill
- dogfood-cliSkill
- enforce-rules-formatSkill
- generate-logoSkill
- hexagonal-architectureSkill
- landing-siteSkill
- pr-to-lint-ruleSkill
- review-docsSkill
- screenshotSkill
- verify-docs-findableSkill
- pr-to-lint-ruleSkill
- adopt-specSkill
- debug-my-harnessSkill
- edit-specSkill
- linter-docsSkill
- strengthenSkill
- test-harnessSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
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 24
Write the answers into the PR body.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Ship a feature (so it ships, not half-ships)
The nine-row definition-of-done already exists — cohesive-feature-delivery in
CLAUDE.md. On 2026-09-02 equivalentDisasters went out tested, gated and
committed, and failed four of those rows unnoticed: unreachable from every
public door, a neighbour's JSDoc silently stolen, no doc, no stability mark.
Prose in a 2000-line file does not fire. This does. Every check below names
the failure it was written for; nothing here is theory.
1. Before writing — three decisions only you can make
| Decision | Criterion |
Which door? (package.json exports subpath) | Free + deterministic → . (src/test.ts). Needs an adapter (files, processes, a harness) → its own subpath served by a composition root in src/entries/<name>.ts (vigiles/rule-proof): a barrel in src/*.ts is app and may not import an adapter. Bills a model → ./eval (paid_ prefix). Hook vocabulary → ./hook. A symbol not re-exported from a door's barrel does not exist to require("vigiles") (failure 1). |
Stable or experimental_? | STABILITY.md: exported + not stable ⇒ @experimental tag AND experimental_ name. One day old with one consumer is not stable. ESLint local/experimental-name holds tag→name; it cannot write the tag (failure 4). |
Where does the WHY live publicly? (document-the-why) | A docs/*.md section beside the claim it changes — for equivalentDisasters, next to "blocks 7/7" in docs/compiled-hooks.md. Not the README (brevity), not this file (contributor tier). |
Write the answers into the PR body. The script below refuses to pass the stability row until you state it.
2. After writing — one command, five checks
npm run build
node .claude/skills/ship-a-feature/scripts/ship-check.mjs <symbol> [--stable "<why it is stable>"]| Check | What it executes | Measured failure it exists for |
| REACHABLE | import()s every exports subpath from the built dist/; the symbol must be a key on ≥ 1 | equivalentDisasters in src/guardrail-check.ts, never in src/test.ts → undefined (1) |
| SURFACE | regenerates api-surface/ (api:report), then READS git diff: any +… (undocumented) fails | inserted between verifyGuardrail's JSDoc and its declaration; the diff said so, nobody read it (2) |
| DOCUMENTED | \b<symbol>\b must occur in docs/**/*.md or README.md | no public home until asked (3) |
| FINDABLE | two pages must name it — or one names it and another LINKS to that section's anchor | documented once, inside a guide about a DIFFERENT feature, with no inbound link (5) |
| MARKED | declaration's JSDoc has @experimental, or the name is prefixed, or you passed --stable "<why>" | exported one day old, no tag, no prefix; the lint rule had nothing to hold (4) |
DOCUMENTED and FINDABLE are different questions. The first asks whether the
capability is written down anywhere. The second asks whether a reader who does
not already know which file to open can get to it. experimental_equivalentDisasters
(since renamed experimental_alternateSpellings) passed the first and failed the second: it was explained once, under its own
heading, inside the guide for a different feature, and nothing linked there — so
the testing guide, where someone asking "does my guard actually block?" looks,
named it zero times. A pointer satisfies FINDABLE; you do not have to repeat the
explanation.
✗ means fix and re-run; the script prints the fix beside each finding. A ✓ MARKED
stable line carries your reason — paste it into the PR body.
3. The gate — on a FROZEN tree, in the foreground, with its real exit code
node .claude/skills/ship-a-feature/scripts/ship-check.mjs --gate # = npm run check, frozenIt hashes the tree before and after and fails if anything changed while the gate ran. Three measured ways a gate lies, all closed by running it this way:
- Editing during the run (5):
npm run checklinted a half-written file and failed on errors that belonged to no state. Do not touch the tree until it exits. - Waiting with
until ! pgrep -f 'check.mjs'(6): the waiting shell's own command line matches the pattern, so it never ends. Twenty minutes reported as "still running". Run gates in the foreground; if you must background one,waitthe PID you spawned. Neverpgrep -fa string that appears in your own command. npx vitest run … | tail -8(7): the pipeline's status istail's. Exit 0 with three tests failing. No pipe after a gate — redirect to a file and read it.
Already mechanized — call it, do not restate it
| Property | Owner (do not duplicate) |
tag @experimental ⇒ experimental_ name | ESLint local/experimental-name (in npm run lint) |
./eval symbols carry paid_ | npm run exports:check (in npm run check) |
| surface drift vs committed snapshot | npm run api:check (in npm run check) |
docs' ts blocks import real exports | npm run docs:check — only once a doc EXISTS (3) |
every CLI verb mentioned in docs/ | src/doc-command-coverage.ts — verbs only, not API symbols |
| the whole command list | npm run check (node scripts/check.mjs --list prints it) |
4. Done means
The five ✓ lines, a ✓ GATE line, and the PR body naming door · stability
decision · doc section. A feature nobody can import is not a feature, and one
nobody can navigate to is not documented.
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.

