name: coverage-report description: Use when the user asks to run, summarize, compare, or standardize SuperContext KG coverage metrics for one repo or a fleet. Builds or uses a snapshot, runs coverage metrics, generates coverage-run.json and coverage-run.md via the repo CLI, and summarizes the highest-value KG coverage gaps without hand-editing numbers.
Coverage Report
Use this skill to produce consistent coverage reports from KG snapshots.
Rules
- Treat
metrics.jsonl,coverage-run.json, andcoverage-run.mdas generated artifacts. - Do not hand-edit metric values or report summaries.
- Prefer deterministic repo commands over ad hoc spreadsheet summaries.
- If a snapshot already has fresh
metrics.jsonl, run only the report step. - If metrics are missing, run
coverage_metricsfirst. - For fleet runs, record
--expected-reposwhenever the expected repo count is known.
Standard Workflow
- Build or locate the snapshot.
For one repo:
python -m source.scripts.build_kg --repo <repo-path> --out <snapshot-dir>
For a fleet:
python -m source.scripts.build_multi_kg --repo <repo-1> --repo <repo-2> --out <snapshot-dir>
- Compute and persist metrics.
python -m source.scripts.coverage_metrics --snapshot <snapshot-dir> --expected-repos <N>
- Generate the stable report.
python -m source.scripts.coverage_report \
--snapshot <snapshot-dir> \
--out docs/evaluation/runs/<run-id> \
--run-id <run-id> \
--tenant <tenant-or-org> \
--expected-repos <N> \
--metric-config source/kg/metrics/config.yaml
- Summarize from
coverage-run.jsonorcoverage-run.md.
Report:
- fleet score
- lowest repo coverage
- weakest dimensions
- worst metrics
- coverage gaps from
coverage_gaps(unsupported languages, uninstrumented stacks, stale/partial coverage) partial/n_areasons and contract flags- narrow next PR recommendation
Verification
When changing the report code or skill:
python -m compileall -q source
python -m unittest tests.metrics.test_report tests.metrics.test_persistence tests.test_packaging_metadata
python -m unittest discover -s tests