| name | evidence-gap-review |
| description | Analyse specgraph verify output and identify what evidence is missing to advance specs to their next state. |
Review the output of specgraph verify (or specgraph verify --json) and produce an actionable gap report.
Procedure
- Run
specgraph verify (or read the provided output).
- For each spec with status
WARN or FAIL:
- Identify the required evidence dimension(s) that are unmet.
- State the current evidence level and what level is needed.
- Suggest the lowest-effort action that would provide the required evidence.
- Group findings by effort: low (annotation), medium (test), high (architecture decision).
Evidence-gap action map
| Required level | Unmet because | Recommended action |
|---|
| E0 | No @spec annotation | Add /** @spec ID @implements Feature */ to the implementation file |
| E1 | No closed Beads issue | Create a Beads issue with --spec-id ID and close it when done |
| E2 | No cross-reference | Add cross_refs: [ID] to a related spec, or add a RELATED section |
| E3 | No test annotation | Add /** @spec ID @verifies Feature */ to a test file |
| E4 | No CI artifact | Configure freshness.artifacts in .specgraph/config.json |
Output format
For each gap, output:
SPEC-ID [WARN|FAIL] dimension=<name> need=<Ex> have=<Ey|none>
→ Action: <one-line description>
Then a prioritised action list (most impactful first).
Rules
- Do not suggest adding waivers unless the requirement is genuinely unreachable.
- Prefer adding annotations over creating new files.
- If multiple specs share the same gap pattern, batch the fix into a single suggestion.
- Check
specgraph explain <SPEC-ID> for detailed claim breakdown when unsure what evidence exists.