repository-quality-score

Run the /repository-quality-score command after audits finish. The skill discovers findings in the current and registered Git worktrees, validates them against the score policy, and invokes the bundled Python calculator to produce score.md, score.json, snapshot.md, and metadata.json.

Missing audits reduce coverage rather than counting as zero scores. The result is labeled provisional, the missing audits are listed by name, and the report explains exactly which evidence is absent so you can rerun those audits.

No. The calculator selects one atomic current-commit result per audit and never merges checks from separate runs or different commits. Results targeting another commit are excluded and those audits must be rerun on the current commit.

Legacy findings using check, gate, or id fields are mapped only through exact unique identifier matches. Ambiguous or unknown identifiers are excluded and reported, and any legacy-derived score is always provisional, never official.

A score is provisional when official conditions are unmet, such as filtered runs, threshold or policy overrides, degraded evidence, skipped checks, a dirty source tree at audit time, or missing audits. The statusReasons field lists the exact cause codes.

The skill never runs or reruns audits, never inspects source code to fill gaps, and never adjusts weights or bands by judgment. It does not claim security or regulatory certification and does not replace the detailed audit reports or human review.