# quality

> Independent verification — evidence, not assertion, before you ship. Installs as an agent skill.

`quality` provides independent evidence that an assembled product meets its functional and
non-functional expectations, and converts that evidence into two separate decisions: may we ship,
and what may we claim. It writes no tests — `frontend`, `backend`, `data`, and `ai` already own
construction-side testing. Its flagship move is naming the oracle behind any check and grading the
result into one of four evidence states (`observed` / `inferred` / `not_tested` / `blocked`) rather
than trusting a green verdict. `SKILL.md` is a faceted router: one primary job (of 11, e.g.
`oracles-and-evidence` ⭐) composes with at most one base surface (of 5: web-app-and-site, service-api,
mobile, data-pipeline, ai-system) plus an additive `surface-regulated` overlay when a conformance
obligation applies — read only the two or three references the request needs, never the whole
18-reference pack. Claude Code is the reference host; it's plain Agent Skills format
(agentskills.io), so it installs on Codex, Cursor, Antigravity, opencode, Grok Build, and Hermes too.

## Install

- npm/skills.sh: `npx skills add gabros20/quality-skill -g -y`
- Clone + installer (per-host targets: claude | codex | cursor | antigravity | opencode | grok | hermes | agents | all): `git clone https://github.com/gabros20/quality-skill && cd quality-skill && ./install.sh codex`
- Manual copy: `cp -R skills/quality ~/.claude/skills/quality`

## Docs

- [README.md](https://raw.githubusercontent.com/gabros20/quality-skill/main/README.md): overview, install, workflow, and repository map.
- [skills/quality/SKILL.md](https://raw.githubusercontent.com/gabros20/quality-skill/main/skills/quality/SKILL.md): runtime router, workflow, artifact contract, and completion rules.
- [docs/installation.md](https://raw.githubusercontent.com/gabros20/quality-skill/main/docs/installation.md): installation and verification by client.
- [docs/usage.md](https://raw.githubusercontent.com/gabros20/quality-skill/main/docs/usage.md): activation boundary, routes, outputs, and completion.
- [docs/recipes.md](https://raw.githubusercontent.com/gabros20/quality-skill/main/docs/recipes.md): representative starting prompts.

## Reference inventory (`skills/quality/references/`)

Jobs (11):
- `verification-strategy-and-risk.md` — risk-ranked altitude allocation, stopping heuristics, "what we deliberately do not verify"
- `oracles-and-evidence.md` ⭐ — **flagship.** naming the oracle; the four evidence states; the "what this evidence does not establish" table
- `functional-verification.md` — journeys, cross-service verification, contract testing's limits
- `accessibility-conformance.md` — conformance vs scan, the machine-testable boundary, VPAT/ACR
- `performance-and-load-verification.md` — field vs lab, percentiles, coordinated omission
- `security-and-privacy-verification.md` — assurance grades, severity precedence conflicts
- `compatibility-and-support-matrix.md` — the matrix, its legitimate sources, and their licence terms
- `ai-behavior-verification.md` — evals as a gate, judge validity, non-determinism, red-teaming
- `suite-and-signal-health.md` — coverage semantics, mutation, flake as an evidence-validity problem
- `defect-lifecycle.md` — ODC classification spine, reproducibility, triage routing
- `release-gate-and-claims.md` — the evidence dossier + adversarial read, the release/claim split

Surfaces (5 + 1 additive) and handoff:
- `surface-web-app-and-site.md` (default) · `surface-service-api.md` · `surface-mobile.md` ·
  `surface-data-pipeline.md` · `surface-ai-system.md`
- `surface-regulated.md` — additive overlay: conformance claims, traceability, evidence retention, verifier independence
- `handoff.md` — what this skill consumes and emits, and the `handoff.yaml` companion's shape

## Optional

- [Visual guide](https://quality-skill.vercel.app): the router, the oracle/evidence flagship, and the full reference map on one page.