← ClaudeAtlas

review-evidence-gapslisted

Review a local audit-preparation workspace and separate supported observations, reported statements, contradictions, missing evidence, and limitations. Use before presenting readiness findings or exporting a human-review bundle.
kotnisofiane-bit/dubsar-agent-skills · ★ 0 · Code & Development · score 68
Install: claude install-skill kotnisofiane-bit/dubsar-agent-skills
# Review Evidence Gaps ## Objective Assess the coherence and support of the preparation, not the compliance or safety of the target system. ## Inputs - an explicit local workspace path containing the five initialized case JSON files; - the approved artifacts referenced by `evidence-index.json`; - recorded scope approval and sensitive-action human review, if provided. ## Workflow 1. Confirm that all five documents use the expected formats and same `case_id`. 2. Verify that approved evidence, inventory sources, inventory items, actions, and artifact paths resolve. 3. Verify indexed file digests and reject forbidden, non-regular, escaping, or credential-bearing artifacts. 4. Require evidence references for supported observations and preserve unsupported statements as `reported` or missing evidence. 5. Separate: - supported observations; - reported statements; - unresolved contradictions; - missing required evidence; - explicit limitations. 6. Use `not_ready` while approval, required preparation, references, contradictions, or required evidence remain unresolved. 7. Use `ready_for_human_review` only when the full preparation is coherent. 8. Run deterministic validation and report every finding and readiness reason. ## Deterministic validation When installed with the complete pack, run `../dubsar-audit-readiness/scripts/validate-audit-workspace.mjs` with an explicit `--root`. Treat its output as structural evidence only. If the script is absent