LegalReviewSkill

---
name: LegalReviewSkill
description: >
  Compare inbound agreements against company templates, highlight risky clauses, and generate revision guidance.
version: 1.2.1
visibility: organization
tags:
  - legal
  - contract
  - compliance
entrypoint:
  command: review_contract
  label: Analyze third-party agreements
---

## When Claude should load this skill
- Invoke when reviewing NDAs, MSAs, SOWs, or other vendor agreements that must align with house templates.
- Use when teams request clause risk scoring, redline suggestions, or approval packets for legal leadership.

## Core workflow
1. Consult `instructions/legal-review-guide.md` for evaluation rubric and escalation thresholds.
2. Run `python scripts/legal_diff.py` to compare inbound language against standards and emit `CLAUSE_DIFF` plus `SUGGESTED_EDIT`.
3. Draft the client-facing response in `templates/risk-summary.md`, attaching tracked changes for review.

## Linked resources
- `templates/master-agreement.docx` — Canonical clause language used as comparison baseline.
- `data/clause_library.json` — Clause variants, fallback options, and commentary from legal ops.
- `instructions/legal-review-guide.md` — Review checklist, fallback approval matrix, and routing guidance.
- `templates/risk-summary.md` — Markdown summary skeleton for surfacing issues to stakeholders.
- `logs/review-register.csv` — Optional audit log for filing completed reviews and owners.

import json
from difflib import SequenceMatcher

with open("data/clause_library.json", encoding="utf-8") as fp:
    library = json.load(fp)

standard = "赔偿条款：如因甲方原因导致损失..."
inbound = "赔偿条款：若乙方提供的服务出现问题..."

ratio = SequenceMatcher(a=standard, b=inbound).ratio()
risk = "high" if ratio < 0.65 else "medium"

print("CLAUSE_DIFF=", {"match_ratio": ratio, "risk_level": risk})
print("SUGGESTED_EDIT=", library["indemnification"]["recommendation"])