| name | adg-evidence-curator |
| description | Select, record, and validate ADG evidence. Use when choosing targeted checks, recording failed commands, linking tests to requirements, writing audit events, registering advisory gaps, or preparing feature/release evidence. |
ADG Evidence Curator
Use this to keep proof truthful and compact.
Decide
- what exact claim the evidence proves;
- the narrowest command or artifact that proves it;
- whether unit, integration, browser, database, security, traceability, or full governance evidence is required;
- whether a failed result needs backlog, audit, or gap registration.
Rules
- Record failed commands instead of hiding them behind a passing summary.
- For a deliverable that renders to a user, a rendered artifact (screenshot or montage) plus
npm run asset:lint on any image assets is the evidence; metric-only evidence is a lower tier and needs a live event under release-class:visual.
- Do not use a narrow check to prove a broad claim.
- Evidence should name commands, paths, artifacts, or audit ids.
- Advisory gaps are allowed, but must remain queryable.
