lexicon verify

Usage

lexicon verify

Description

Runs the full verification pipeline:

1. Load gates — Reads specs/gates.toml (falls back to the default model if not found)
2. Run all gates — Executes each gate command as a subprocess via sh -c in the repository root
3. Load scoring model — Reads specs/scoring/model.toml if it exists
4. Compute score — Maps gate results to scoring dimensions and computes the weighted score
5. Write audit record — Records the verification run in .lexicon/audit/

Output

Running verification
────────────────────────
Gates
 ✓ fmt (42ms)
 ✓ clippy (1830ms)
 ✗ unit-tests (2105ms)
 ⊘ doc-tests (0ms)

Score
 Total: 65.0% (WARN)
  correctness: 100% — fmt: passed in 42ms
  conformance-coverage: 50% — no gate configured
  behavior-pass-rate: 50% — no gate configured
  lint-quality: 100% — clippy: passed in 1830ms
  doc-completeness: 100% — advisory
  panic-safety: 0% — unit-tests: FAILED
────────────────────────
Some gates failed

Gate Outcome Icons

Icon	Meaning
✓	Pass
✗	Fail
⊘	Skip
!	Error

Gate-to-Dimension Mapping

During scoring, gate results are mapped to dimensions by matching gate.id to dimension.id:

• If a gate matches and passed: dimension value = 1.0
• If a gate matches and failed: dimension value = 0.0
• If no gate matches an advisory dimension: value = 1.0 (assumed pass)
• If no gate matches a non-advisory dimension: value = 0.5 (partial credit)

Audit Record

Each verification run writes a JSON audit record to .lexicon/audit/ with:

• Action: verify_run
• Actor: system
• Summary: gate count and pass count
• Whether all gates passed
• Final score (if scoring model exists)

Exit Behavior

The command currently always exits successfully. Check the output for gate failures and score verdicts. In CI, you may want to combine this with a script that checks the output.