consistency-audit
Skill1 starsChanged 36 days agoApache-2.0View on GitHub
Read-only consistency audit of a PR or diff against CURRENT, accepted ADRs, and project locks before merge. Use when CoS/QC asks for MERGE_OK, a consistency grill, or docs/impl PR quality check. Not for inventing product and not a substitute for maintainer-only gan-critique failure-mode review (linked from snowshoe-maintainers when present).
4 other projects use a version of this.
Adopted by
Most recently taken up by
- Sunnyshinnny776/pm-agentic-cookbooktheir own version, first seen Oct 20, 2026
- lcandy2/agent-skillstheir own version, first seen Oct 5, 2026
- wanshuiyin/Anti-Autoresearchtheir own version, first seen Jun 27, 2026
- 45ck/hci-review-skilltheir own version, first seen Mar 28, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- wanshuiyin/Anti-Autoresearch★ 180same name
- 45ck/hci-review-skill★ 17same name
- lcandy2/agent-skills★ 4same name
- Sunnyshinnny776/pm-agentic-cookbook★ 2same 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 repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .cursor/rules/agent-comms.mdcCursor rule
- .cursor/rules/project.mdcCursor rule
- brain-protocolSkill
- grill-canonSkill
- grillingSkill
- grill-meSkill
- snowshoe-scratchSkill
- snowshoeSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 4 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
4
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 38
Do not treat EXPERIMENTS.md or docs/evidence/ as truth.
- Distinctive: line 72
Prefer facts from CURRENT/accepted ADRs/code over the PR body's claims
- Distinctive: line 73
Do not invent product direction
- Distinctive: line 74
Do not open a smell-only rewrite epic
- Distinctive: line 75
If GAN already ran: do not re-litigate Closed items.
“Common” is measured: kept by at least 6 in 10 of the 3 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
Consistency audit
You are a read-only auditor. Goal: does this change stay consistent with truth and locks? You do not author the PR. You do not merge.
Distinct from:
grill-canon— audits the brain for staleness (CURRENT/ADRs themselves)gan-critique— optional maintainer discriminator (siblingsnowshoe-maintainers; may be absent)grilling— interview rounds for undecided product
This skill is the merge QC companion: CI is mechanical; you check narrative and lock alignment.
Inputs
Require (or look up):
- PR number / branch / head SHA, or an explicit path list + base..head
- Kind hint:
docs|implementation|mixed
Read order
- PR title/body + file list + diff (prefer remote
gh/ GitHub tools; no need to invent) docs/brain/CURRENT.md+ every accepted ADR the diff touches or implies.cursor/rules/project.mdclocks- Related notes only if CURRENT cites them
- For implementation: tests/CI status on the head if available
Do not treat EXPERIMENTS.md or docs/evidence/** as truth.
Checklist
Report against each item with evidence (path / SHA / quote):
- CURRENT ↔ ADR — if behavior or product claim changed, is there an ADR and CURRENT update in the same change (or already true on main)?
- Locks — honor always-on
.cursor/rules/project.mdc+ CURRENT + accepted ADRs. Do not restate those locks here. - Skill bodies (if
.cursor/skills/**touched) — role-generic; locks live in project.mdc / CURRENT, not enumerated in the skill - Scope — diff matches stated slice; no silent drive-by
- CI — brain-docs / app checks green or explained
- Draft hygiene — undraft only when ready; auditor does not undraft/merge unless asked
Verdict shape
Output exactly:
## Consistency audit — <PR or ref> @ <sha>
**Verdict:** MERGE_OK | MERGE_BLOCKED | NEEDS_PRODUCT
**Blockers** — each: failure vs canon/lock, where, what must change (or product letter options)
**Smells** — graded `should` / `note` (do not block alone)
**Aligned** — short bullets of what already matches CURRENT/ADRs
**Out of scope / deferred** — one line eachMERGE_OK— no blockers; smells optionalMERGE_BLOCKED— canon/lock/acceptance miss; generator or author must patchNEEDS_PRODUCT— cannot decide from canon; lettered options for the user
Rules
- Prefer facts from CURRENT/accepted ADRs/code over the PR body's claims
- Do not invent product direction
- Do not open a smell-only rewrite epic
- If GAN already ran: do not re-litigate Closed items; only raise new consistency misses
- User-facing chat defaults to English (see
.cursor/rules/agent-comms.mdc). File patches and proposed docs stay English unless the artifact itself is user-facing prose in another language.
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.

