Scoring model

Every finding is weighted by severity. The overall grade is a function of the weighted pass rate, with an extra penalty for CRITICAL failures so a single critical issue cannot be masked by many low-severity passes.

Weights

Severity	Weight
CRITICAL	20
HIGH	10
MEDIUM	5
LOW	2
INFO	0

INFO findings are informational and do not move the score.

Formula

base          = (sum of weights for passing checks) / (sum of all weights) * 100
critical_pen  = 5 points per CRITICAL failure
score         = clamp(round(base - critical_pen), 0, 100)

If every finding is INFO (or there are none), the score is 100.

Grade bands

Score	Grade
≥ 90	A
≥ 75	B
≥ 60	C
< 60	D

Exit codes

The full exit-code table is documented in usage.md. The default gate is --fail-on CRITICAL: one CRITICAL finding in the effective set (after baseline + ignore filtering) fails CI. The grade is not the default gate criterion; use --min-grade to gate on it explicitly. See ci_gate.md for the full gate contract.

The implementation lives in pipeline_check/core/scorer.py and pipeline_check/core/gate.py.