AgentSkillsCN

Legal Warroom

法律作战室

SKILL.md

LEGAL WARROOM — Team Coordinator

Purpose

Coordinates the 10 legal specialist agents (L01-L10) to produce a complete, export-ready legal work product from raw documents. Manages the pipeline: intake → timeline → evidence → claims → strategy → drafts → QA → export.

Triggers

  • •[LEGAL_RUN] command
  • •Any task involving: Rechtsstreit, Klage, Beweis, Anwalt, Gericht, Verteidigung, Vergleich
  • •New files in data/inbox/ tagged as legal
  • •User requests for legal analysis, timeline, evidence mapping

Inputs

  • •Raw documents in data/inbox/ (PDFs, emails, screenshots, contracts)
  • •Prior legal outputs in legal/ subdirectories
  • •agents.json entries for L01-L10

Outputs (mandatory deliverables)

  1. •legal/timeline/TIMELINE.md — Master chronology
  2. •legal/evidence/EVIDENCE_MAP.md — Evidence index with chain of custody
  3. •legal/claims/CLAIM_MATRIX.md — Claims, defenses, evidence links, risk scores
  4. •legal/strategy/OPPONENT_ANALYSIS.md — Opposing party behavior patterns
  5. •legal/strategy/RISK_REPORT.md — Best/worst case scenarios with probabilities
  6. •legal/strategy/SETTLEMENT_PLAN.md — Negotiation strategy + BATNA
  7. •legal/drafts/ — Letters, motions, memos as needed
  8. •legal/memos/CONSISTENCY_REPORT.md — Cross-check timeline vs evidence vs claims
  9. •legal/memos/EXEC_BRIEF.md — 1-2 page executive summary

Playbook

Phase 1: Data Ingest (D01 + D02 + D03)

  1. •Inventory all files in data/inbox/
  2. •OCR any image-based documents
  3. •Normalize naming: YYYY-MM-DD_TYPE_DESCRIPTION.ext
  4. •Move processed files to data/processed/

Phase 2: Timeline (L01 - Legal_Timeline)

  1. •Extract all dates, events, actors from processed documents
  2. •Build chronological timeline with source references
  3. •Flag gaps: periods with no documentation
  4. •Output: legal/timeline/TIMELINE.md

Phase 3: Evidence Mapping (L02 - Legal_EvidenceMapper)

  1. •For each document, identify: what it proves, who it involves, relevance level
  2. •Create evidence chain: Document → Fact → Legal Claim
  3. •Apply exhibit naming: EX-001, EX-002, ...
  4. •Output: legal/evidence/EVIDENCE_MAP.md

Phase 4: Claims Matrix (L03 - Legal_ClaimsMatrix)

  1. •List all potential claims (yours AND opponent's)
  2. •For each claim: supporting evidence, risk level (1-5), strength assessment
  3. •Map defenses and counter-arguments
  4. •Output: legal/claims/CLAIM_MATRIX.md

Phase 5: Strategy (L05 + L06 + L07)

  1. •Opponent analysis: patterns, likely moves, weaknesses
  2. •Risk assessment: best/worst/likely scenarios with probability %
  3. •Settlement strategy: your BATNA, their BATNA, zone of agreement
  4. •Outputs in legal/strategy/

Phase 6: Drafts (L08 - Legal_Drafting)

  1. •Draft any requested documents (letters, motions, responses)
  2. •Reference timeline + evidence map in footnotes
  3. •Output in legal/drafts/

Phase 7: QA (L09 + L10)

  1. •Cross-check: Do all timeline events have evidence? Do all claims have support?
  2. •Consistency check: No contradictions between documents
  3. •Executive brief: 1-2 page summary for decision-maker
  4. •Outputs: legal/memos/CONSISTENCY_REPORT.md, legal/memos/EXEC_BRIEF.md

Safety & Quality Checks

  • •Every factual claim MUST reference source document + page/section
  • •Mark uncertain items as [UNVERIFIED] with confidence level
  • •This system does NOT replace legal counsel — note this in every output header
  • •Never fabricate evidence or dates
  • •Flag any document that might be privileged (attorney-client communication)

Exhibit Naming Convention

code
EX-001  First exhibit (chronological order of relevance)
EX-001a Supplement to EX-001
EX-002  Second exhibit

YAML Header Template (for all legal outputs)

yaml
---
title: <Document Title>
agent: <L01-L10 Agent Name>
team: Legal
created_at: <ISO8601>
inputs: [<source files used>]
confidence: <low|medium|high>
case_ref: <Case identifier if applicable>
---