@language_reasoning/legal_clause_comparison
no model is authoritative on binding legal decisions; always human gate
I need to compare termination clauses across two SaaS contracts
activate@language_reasoning/legal_clause_comparison
Opus 4.8 correctly activates exception conditions and scopes applicability. Sonnet 4.6 for triage. Attorney review required for any binding decision.
NDA triage, first-pass review, volume task, non-binding
claude/sonnet-4.6multi-party contracts, exception activation, compliance analysis
claude/opus-4.8any binding legal decision
human (attorney)| ID | Relation | Actor | Result | Output |
|---|---|---|---|---|
| CT-D8-001 | activate@exception_condition | claude/opus-4.8 | accepted | activates unless/provided-that conditions in contract clauses |
| CT-D8-002 | scope@applicability | claude/opus-4.8 | accepted | bounds clause applicability to stated jurisdiction and conditions |
| CT-D8-003 | identify@clause_type | claude/sonnet-4.6 | accepted | reliable for first-pass clause identification and NDA triage |
| CT-D8-004 | validate@jurisdiction | claude/opus-4.8 | bounded | useful for scoping but not authoritative; requires attorney review |
Opus 4.8 Legal Agent Benchmark: highest recorded, first to break 10% all-pass
E-GLOBAL-008Opus 4.6 scored 90.2% on Harvey BigLaw Bench
E-GLOBAL-011Sonnet 4.6 suitable for NDA triage and first-pass review at 5x lower cost
The Legal Prompts, Feb 2026No AI model is a reliable final authority on binding legal decisions
professional standards + benchmark design