harness-capstone-audit

Run a structured audit that scores nine dimensions: entry, context, state, feedback, evaluation, runtime observability, benchmark, cleanup, and plugin readiness. Each dimension is graded from file and command evidence, producing an overall 0-10 score and the next best improvement.

The audit checks entry (root instructions, command routing), context docs, state management (WIP=1, progress), feedback (startup path, verification command), Judge evaluation profiles, runtime observability, benchmarks, cleanup protocol, and plugin readiness.

Scores are capped by missing fundamentals: a missing startup path caps the score at 6, missing state or Judge evaluation caps it at 7, and unsafe auto-mutating plugins cap it at 8. Fix the flagged gap to raise the ceiling.

Baseline readiness means the minimal harness elements exist, such as entry instructions, state tracking, and a verification path. Capstone maturity (9+) requires benchmarks, runtime observability, cleanup protocols, and safe plugin behavior on top of that baseline.

No. The audit rules require scoring from file and command evidence, not optimism or planned capabilities. Configured but unverified features must not be treated as accepted without corresponding evidence.