quality-gate
Skill5 starsChanged 6 months agoApache-2.0View on GitHub
Pre-delivery review of a deliverable: structural blockers first, then substance. Use before anything leaves the building — external email, published content, submitted work.
46 other projects use a version of this.
Adopted by
Most recently taken up by
- MarchTalk/ai-content-skillstheir own version, first seen Nov 16, 2026
- Fili999/ai-hubtheir own version, first seen Nov 15, 2026
- shimo4228/writingtheir own version, first seen Nov 15, 2026
- erraggy/oastoolstheir own version, first seen Nov 15, 2026
- shimo4228/zenn-contenttheir own version, first seen Nov 15, 2026
- jodyjdc/ultimate-best-seller-studiotheir own version, first seen Nov 8, 2026
Used with
Projects with it also use deploy.
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.
VercelSkilldeploy
Deploy applications to Vercel. Use when the user says "deploy", "deploy to Vercel", "push to production", "deploy my app", or "go live".
Also in 3 projects that use this
Other versions46 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 46
- RapierCraftStudios/ForgeDock★ 119same name
- microsoft/webui★ 102same name
- impecablemee/gtm-mcp★ 66same name
- microsoft/devsquad-copilot★ 50same name
- Romandredan/1c-quality-gate★ 50same name
- TerryFYL/ai-research-army★ 9same name
- Felpix-Studios/social-science-research★ 8same name
- BerryKuipers/claude-code-toolkit★ 7same name
- erraggy/oastools★ 7same name
- MarchTalk/ai-content-skills★ 4same name
36 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 repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- db-healthSkill
- dreamSkill
- forum-pulseSkill
- genesisSkill
- grill-meSkill
- handoffSkill
- hireSkill
- inbox-processSkill
- kaizenSkill
- quarterly-reviewSkill
- session-endSkill
- session-startSkill
- sync-upstreamSkill
- twin-updateSkill
- voice-gateSkill
- weekly-reviewSkill
- write-a-skillSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 4 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 36 edited versions we read keep these.
4
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 18
Run the structural pass first.
- Distinctive: line 31
Not every deliverable needs the same score.
- Distinctive: line 47
Run each check and report PASS/FAIL with the evidence.
- Distinctive: line 134
If the same item fails repeatedly across deliverables, the convention may be the problem rather than the author.
“Common” is measured: kept by at least 6 in 10 of the 36 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
Quality Gate — pre-delivery review
The user-invoked counterpart to .claude/protocols/quality-gate.md. The protocol
decides what gets gated; this skill is how one item is reviewed.
Two tiers. Structural items are blocking — they are mechanical, cheap, and have no judgement in them. Substance items are warnings — they need a reader, and they are where the review actually earns its cost.
Run the structural pass first. A reviewer who spends their attention on a missing caption has none left for the argument.
Inputs
- File path — passed as an argument, or asked for if missing.
- The original briefing — MANDATORY. Read it before the draft. Content the requester asked for verbatim must not be reported as invented, and a reviewer who flags the requester's own words loses the requester's trust permanently.
- Your voice rules, if you keep any:
.claude/rules/voice-banlist.md.
Step 0 — Set the bar
Not every deliverable needs the same score. Establish the type first, then the minimum, then review against it:
| Deliverable | Blocking items that must pass |
| Internal memo / working note | Structural items only, loosely |
| Report for a named audience | All structural, no open warnings on substance |
| Anything sent to a third party | All structural, all substance |
| Anything published under a brand | All structural, all substance, plus a voice pass |
| Draft explicitly marked as a draft | The gate does not apply |
A single universal bar means either everything is over-reviewed or the bar is set where the cheapest item can clear it.
Step 1 — Structural pass (BLOCKING)
Run each check and report PASS/FAIL with the evidence. "Item 4 failed" is not a finding; "line 82 repeats line 81 verbatim" is.
- Naming — a predictable, sortable filename: date, subject, descriptor.
- Metadata — title, author, date present in whatever the format supports (YAML front matter, document properties, a title slide).
- Repeated adjacent paragraphs — the classic generation artefact:
awk 'NR>1 && length($0)>10 && $0==prev {print NR": "$0} {prev=$0}' <FILE> - Figure flow — every image has its own caption; no bare image immediately after a heading; the rhythm is heading → text → figure → text.
- Convention consistency — callouts, symbols and colours applied consistently, or not at all. Half-applied is worse than absent.
- Language consistency — one language throughout, correct diacritics, one font. A document that drops its accent marks halfway through was written in two sittings and reads like it.
- Cross-references resolve — internal links land, every citation has an entry, figure and table numbers are right, external URLs spot-checked.
- Tables and structure — equal column counts per row, consistent list indentation, code blocks with a declared language, nothing cut off at a margin.
- No generation markers — the sycophantic openers, the hedging phrases, the
throat-clearing transitions. If you keep a banlist, run
/voice-gate <file>here instead of grepping by hand, and let this item inherit its verdict.
Step 2 — Substance pass (WARNINGS)
- Density — long sentences carrying little information, repetition, adjectives standing in for facts. Or the reverse: a wall with no headings.
- Actionability — does it answer "so what happens now?" An analysis needs recommendations with an owner. A memo needs the decision it is asking for. A manual needs steps someone can actually follow.
- Fidelity — every claim traceable to a source or to the briefing; the framing matches the real context rather than a generic version of it; nothing invented to fill a gap the author could have left open.
Step 3 — Score and verdict
=== QUALITY GATE — <filename> ===
Type: <type>
[1] Naming PASS/FAIL
[2] Metadata PASS/FAIL
[3] Repeated paras PASS/FAIL
[4] Figure flow PASS/FAIL
[5] Conventions PASS/FAIL
[6] Language PASS/FAIL
[7] Cross-refs PASS/FAIL
[8] Tables/struct PASS/FAIL
[9] Voice markers PASS/FAIL
---
[10] Density PASS/FAIL (warning)
[11] Actionability PASS/FAIL (warning)
[12] Fidelity PASS/FAIL (warning)
BLOCKERS: <failed structural items, or "none">
WARNINGS: <failed substance items>
VERDICT:
DELIVER — clears the bar for this type
DELIVER WITH NOTES — clears the bar, N warnings worth addressing
DO NOT DELIVER — N blocker(s), fix and re-run
NEXT ACTIONS:
1. <specific fix, with the line it applies to>Step 4 — Evidence for every FAIL
For each failure: the exact line or section, what is wrong with it, and the suggested replacement. A verdict without evidence is an opinion, and the author is entitled to disagree with it.
Step 5 — Log the outcome
INSERT INTO activity_history (actor_id, action, entity_type, summary, metadata, occurred_at)
VALUES (1, 'quality_gate_run', 'document',
'QG <filename>: <verdict>',
json_object('file','<path>','blockers',<n>,'warnings',<n>,'verdict','<verdict>'),
strftime('%Y-%m-%dT%H:%M:%SZ','now'));Notes
- This skill diagnoses; it does not auto-fix. The author decides what to change.
- If the same item fails repeatedly across deliverables, the convention may be the problem rather than the author. Say so.
- Compose with
/voice-gate: run that first for A/B/C-class work, so the mechanical voice violations are gone before a reviewer's attention is spent.
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.

