artefacts-create-auditlisted

Audit a topic and render the findings as a self-contained, actionable HTML page: findings grouped by delivery status, ranked by severity, each verified before it ships. Uses the visual-explainer plugin's rendering patterns. The output is the audit artefact this skill was distilled from: a masthead + KPI row, findings grouped into collapsible **To do / In progress / Done** sections, severity shown on both a chip and the colour of each finding number, a severity/type filter bar, and a refuted section that keeps the audit falsifiable. ## Step 1 — Interpret `$ARGUMENTS` (auto-detect) - **Empty** — ask the user what to audit, then stop. Do not invent a topic. - **A path to a `.json` file** (matches something like `*.json` and the file exists) — **render-only mode**. Load it as the findings dataset and skip to Step 3. Expected shape is documented in Step 2. - **Anything else** — treat the whole string as the **topic** to audit (a roadmap, a design doc, a subsystem, a PR, a migration). Proceed to Step 2. ## Step 2 — Gather findings (topic mode) Analyse the topic and produce a findings dataset. Read the relevant docs and code first; for a broad or uncertain scope, fan out parallel read-only searches (an `Explore` subagent per area) before drawing conclusions. For a genuinely large or adversarial audit, a Workflow (multi-lens find → dedupe → adversarial verify → completeness critic) produces a far stronger set; use it when the topic warrants the cost. Each finding must carry: -