$npx -y skills add WILLOSCAR/research-units-pipeline-skills --skill evidence-auditorUse when paper-review needs a claim-by-claim evidence gap report grounded in an extracted claim ledger. Trigger: evidence audit, missing evidence, unsupported claims, 审稿证据审计, 证据缺口. Use when: paper-review 流程中,需要逐条检查 claim 的证据链、缺 baseline、评测薄弱点。 Skip if: 缺少 claims 输
| 1 | # Evidence Auditor |
| 2 | |
| 3 | Transforms a claim ledger into a gap report for `paper-review`. |
| 4 | |
| 5 | ## Input |
| 6 | |
| 7 | - `output/CLAIMS.md` |
| 8 | |
| 9 | ## Output |
| 10 | |
| 11 | - `output/MISSING_EVIDENCE.md` |
| 12 | - `output/EVIDENCE_AUDIT.jsonl` (`review-evidence-gap.v1`) |
| 13 | |
| 14 | ## Contract |
| 15 | |
| 16 | Each gap block should include: |
| 17 | - claim reference |
| 18 | - what evidence is already present |
| 19 | - what is missing or weak |
| 20 | - minimal fix |
| 21 | - severity |
| 22 | |
| 23 | Each JSONL record must retain `claim_id` and a stable `gap_id` so the final |
| 24 | review can cite the exact failure rather than paraphrasing an unlocated concern. |
| 25 | |
| 26 | ## Script boundary |
| 27 | |
| 28 | `scripts/run.py` should: |
| 29 | - iterate existing claims only |
| 30 | - classify the likely evidence risk |
| 31 | - write actionable, bounded gap items |
| 32 | |
| 33 | It should not invent new claims or rewrite the manuscript. |
| 34 | |
| 35 | ## Acceptance |
| 36 | |
| 37 | - every claim has a corresponding evidence note or gap item |
| 38 | - minimal fixes are actionable and concrete |
| 39 | - structured records join back to every `claim_id` |
| 40 | |
| 41 | ## Non-goals |
| 42 | |
| 43 | - manuscript rewriting |
| 44 | - novelty assessment |
| 45 | - recommendation writing |