Coverage Integration
Read when changed files have CI coverage or a coverage gap affects risk. Discover the actual report path and parse its format: LCOV, Cobertura, or Istanbul. A file-level summary alone cannot prove changed-line coverage; report that evidence gap rather than substituting whole-file coverage.
Decision Thresholds
These are Guardian routing/scoring thresholds, not universal test-quality guarantees.
| Signal | Classification |
|---|---|
Changed-line coverage < 0.50 |
Critical changed-line signal |
Auth/security coverage < 80%; new payment logic < 90%; changed code 0% |
Critical gap |
Core business logic < 70%; hotspot < 60%; regression risk > 5% |
High-priority gap |
Route to Radar on any of:
- High-risk file with
coverage_gap > 0.40. - Hotspot with coverage
< 0.50, or critical file with no tests. - Coverage regression
> 5%with high risk, orregression_risk > 0.70.
Score
The Test Score component weights line coverage 30%, branch coverage 25%, changed-line coverage 25%, test quality 10%, and coverage delta 10%. Its overall PR-quality weight is 15%.
Handoff and Output
GUARDIAN_TO_RADAR_HANDOFF (Coverage) includes PR / Branch, Reason, Critical files with measured coverage/evidence, and Requested action: focused tests, closure of changed-line gaps, and residual-risk report.
Use Coverage Gap Analysis with Critical Gaps (Must Fix) and High Priority Gaps; use Coverage Integration Report with PR Coverage Summary and Test Score Breakdown. Include only measured results; absent coverage stays unknown.
AUTORUN may parse, correlate, and report. Pause when a critical gap combines with another blocking condition or inconsistent data invalidates scoring. Missing data follows reference/autorun-mode.md partial-result handling.