Apply Review

Apply Review

Apply confirmed decisions to a new map. Preserve the source artifact.

Follow ../../references/interaction-contract.md.

Process

1. Read references/decision-contract.md.
2. Collect all independent blocking decisions into one compact batch.
3. Require a reason and evidence reference for each distinct value, exclusion, or shared-scope disposition; reuse one evidence reference across its full scope.
4. For byte order, require the exact evidence file and matching sample identity.
5. Run python3 <skill-dir>/scripts/run.py --map <draft.json> --decisions <decisions.json> [--evidence <evidence.json>] --output <reviewed.json>.
6. Run once for the complete batch and report applied decisions, exclusions, audit data, and only the remaining holds.

For a complete supplied decision file, let the wrapper validate its map binding and evidence before investigating implementation details. Inspect the resulting validation or audit record once. Repeated hashing and scratch validators are unnecessary unless a concrete discrepancy remains; never repair a stale decision by rebinding it to a newer map.

Output files

• reviewed.json - Use this new reviewed map. It contains the accepted changes, exclusions, remaining holds, and the evidence trail. The original draft is unchanged.

Completion requires a new map whose approval state matches its remaining holds.

Stop

• Stop without a reason and evidence reference for each distinct decision.
• Do not modify the original draft; write a new map.
• Keep write-only points out of the read map.
• Rebuild every stale read plan.

Handoff

• Holds remain: suggest review-evidence.
• The map changed and is ready: suggest plan-reads.

Keep write-only points out of the read map. Rebuild every stale read plan.