/ratchet:repo-audit — evidence-based feature discovery

A codebase is a claim about what it does. This command verifies the claim by evidence. It discovers user-facing features from the code itself — not from the README, not from memory — and marks the gaps between what exists and what is tested.

Step 0 — Load state and ground truth

ratchet snapshot repo

Then read for real: routes, entrypoints, handlers, UI components, API definitions, config, and existing tests. Use Grep/Glob to find route tables, exported handlers, and test files. Do not infer a feature you cannot point to in code.

Procedure

For each user-facing feature discovered by evidence, capture:

• Feature ID and name
• User story — who does what, and why
• Evidence in code — file:line where it lives
• Expected behavior — what it should do
• Edge cases — the inputs that stress it
• Test cases — what would verify it
• Dependencies — what it relies on
• Known assumptions — what it takes for granted
• Current risk level — critical / high / medium / low

Then identify the gaps:

1. Screens / routes / workflows not yet documented.
2. APIs or state transitions that need tests.
3. The highest-risk feature — most user-facing × least tested.
4. The first test suite to write.
5. The smallest fix likely to improve user trust.

Rules

• Do not declare coverage complete unless every route, screen, workflow, and API interaction has evidence. Missing evidence is a finding, not a rounding error.
• A feature with no test is a defect waiting to happen — record it.

Output contract

Return the feature inventory as a table, then the five gap findings. Then serialize:

ratchet ledger create
ratchet ledger update features '{"name":"...","area":"...","routes":"...","status":"discovered"}'
ratchet defect add '{"severity":"high","summary":"<highest-risk feature> has no test coverage"}'

Next: /ratchet:qa-ledger to formalize the canonical ledger, then /ratchet:verify.