Check
A curated collection of autonomous AI agent skills powered by Vercel's `skills` npm cli package. Specialized tools for productivity, document management, and custom LLM workflows.
npx -y skills add arendon1/agent-skills --skill checkAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Read-only drift detector. Diffs the plan artifacts against the current code and reports violations grouped by severity. Writes nothing — suggests remedies but never invokes them. Use when checking drift, auditing the spec, verifying invariants hold, or when the user says "check drift", "audit the spec", "does the code still match", "spec vs code", "is the plan still accurate".
SKILL.md
3.9 KB, as published. Nobody here has run it
check — drift report
Pure diagnostic. Reports violations grouped by severity. Writes nothing. The user decides the remedy.
Plan artifacts drifting silently from code is the #1 failure mode. check is the
detector. Run it after each build and before each ship — drift caught here is a
diff; drift caught in prod is a LESSONS.md bug row.
OWNERSHIP
Owns: the drift report (delivered inline — does NOT write a file).
Reads: PLAN.md, SPEC.md, ARD.md (whatever exists), CONTEXT.md, the codebase.
MUST NOT write anything. NEVER invokes fix skills — only suggests them.
WHEN (user-invoked)
- After a
buildcompletes a task or a plan. - Before a ship / merge / PR.
- User asks "check drift", "audit the spec", "does the code still match", "is the plan still accurate".
- Invoked by the
verifydiscipline as the final plan-level drift check.
STEP 0 — LOAD
- Resolve the active plan per §6. If none, "no plan, nothing to check." Stop.
- Read whatever exists:
PRD.md,ARD.md,SPEC.md,PLAN.md,CONTEXT.md. - Parse invocation args:
§V-> check invariants only (default)§I-> check interfaces§T-> audit task status vs code--all-> all three
CHECK §V — invariants
For each V<n> in SPEC.md/ARD.md:
- Translate the invariant into a verifiable claim about the code.
- Grep / read the relevant files.
- Classify:
- HOLD — the code satisfies the invariant.
- VIOLATE — the code breaks the invariant (cite file:line).
- UNVERIFIABLE — no test covers the invariant's path.
- Record the address + file:line evidence.
CHECK §I — interfaces
For each §I item:
- Locate the implementation.
- Classify:
- MATCH — shape in code = shape in spec.
- DRIFT — impl exists, shape differs (cite file:line).
- MISSING — impl absent.
- EXTRA — code exposes a surface not in
§I.
CHECK §T — tasks
For each T<n> in PLAN.md:
- If
[x]: verify the claimed work is present in the code. If not -> STALE. - If
~: note as in-progress. - If
[ ]: note as pending.
REPORT
Caveman. Grouped by severity. Example:
## §V drift
V2 VIOLATE: auth/mw.go:47 uses `<` not `<=`. see LESSONS.md B.1.
V5 UNVERIFIABLE: no test covers every req path.
## §I drift
I.api DRIFT: POST /x returns `{result}` not `{id}`. route.go:112.
I.cmd MISSING: `foo bar` absent from cli/*.go.
## §T drift
T3 STALE: status [x], no middleware file exists.
## summary
2 violate. 1 missing. 1 stale. 1 unverifiable.
next: spec skill with `bug:` or fix code at cited lines.
REMEDY HINTS (not actions)
End the report with a one-line hint per class. NEVER invoke the fixes.
- VIOLATE / DRIFT -> invoke the
specskill withbug: <V.n>or fix the code. - MISSING -> invoke the
buildskill on§T.nif a task exists; else thespecskill withamend §T. - STALE -> the
specskill withamend §Tto uncheck. - EXTRA -> the
specskill withamend §Ito document, or delete the code. - UNVERIFIABLE -> the
tdddiscipline to write the missing test.
BOUNDARIES
- MUST write nothing. No artifact edits. No code edits.
- MUST NOT invoke fix skills. Report only.
- MUST cite file:line for every VIOLATE / DRIFT / STALE finding.
- MUST run on the main thread — no subagents.
- MUST give a binary verdict per item: holds or drifts. No scores, no grades.
- MUST use
cavemanfor the report.