Knuckles-Team Legal Compliance Review Workflow
CONCEPT:KG-2.12 — Company Operations Domain
Coordinates legal compliance review including statute research, regulatory tracking, and compliance validation against the KG ontology_legal.ttl.
Steps
Step 0: Legal Research
Agent: legal-research-agent
Tools: graph_search, graph_query, sx_search
Search US law databases via ScholarX and web-search. Query KG for existing case law, statutes, and precedent relevant to the compliance matter.
Step 1: Contract Analysis
Agent: contract-analyzer
Tools: graph_analyze, document_tools
Parse contracts using document-tools, extract clauses, identify risk factors. Cross-reference against ontology_legal.ttl ContractClause classes.
Step 2: Compliance Validation
Agent: compliance-checker
Tools: graph_query, graph_write
Validate findings against applicable FederalStatute, StateStatute, and MunicipalCode entries in the KG. Flag non-compliance issues.
Step 3: KG Persistence
Agent: graph-os
Persist findings as LegalMatter + CaseLaw + RegulatoryFiling nodes in KG. Create edges linking to applicable statutes via governedByStatute.
Step 4: Human Escalation
Route all binding legal determinations to human oversight. Generate a summary report with recommendations for human review.
Output
- LegalMatter node in KG with status + findings
- Compliance score (0.0–1.0)
- Risk assessment with recommended actions
- Escalation report for human review
Human Oversight Required
✅ All final legal decisions require human approval.
Execution
Run this workflow as a dependency-ordered DAG. Steps with no unmet depends_on run in parallel; dependents run after their prerequisites complete.
- Run first (in parallel): Step 0 — Legal Research; Step 1 — Contract Analysis; Step 2 — Compliance Validation; Step 3 — KG Persistence; Step 4 — Human Escalation
Execution: If graph-os is reachable, offload the whole DAG via graph_orchestrate action=execute_workflow (or the kg-delegate skill) for true parallel/swarm execution. Otherwise execute the steps natively in dependency order: run steps with no unmet depends_on in parallel, then their dependents.