Route: Legal Research

What this skill does

Pick the right LLM for legal research and analysis: issue-spotting, rule application, case and statute analysis, memos and multi-step agentic research. Grounded in mid-2026 benchmarks (Vals AI LegalBench across 124 models, Harvey Legal Agent Benchmark for agentic work).

• Model Routing
• Research
• Benchmarks

Where it runs, and how to install it

Agent Skills are a format, not a product feature: a folder with a SKILL.md and whatever files it needs. Nothing inside one is model-specific.

Claude

Drop the skill folder into ~/.claude/skills/, or upload it in Settings → Capabilities. Claude loads it when the description matches what you asked for.

Claude Code

Same folder, per-project instead: .claude/skills/ inside the repository, so the skill travels with the work rather than with the machine.

Any agent that reads SKILL.md

A skill is a folder with a SKILL.md and whatever files it needs. Nothing in the format is model-specific, so a runtime that reads the file reads the skill.

More from Stephane Boghossian

• Connecticut Divorce PlannerClaude skill that turns Claude into a Connecticut-specific divorce planner: nine operating modes from pre-flight intake to post-judgment modification, modeled on Untangle.us's feature surface and grounded in C.G.S. Title 46b, Practice Book Chapter 25, and the 2026-08-01 CCSG schedule. Covers eligibility triage (nonadversarial under § 46b-44a vs standard), financial affidavit (JD-FM-6), child supp
• Fintech Agreement DraftingDrafting copilot for complex, multi-pillar regulated fintech agreements, where a licensed payment-services provider engages a counterparty across several service lines (agent cash-in/cash-out, QR payments, wallet e-payments, marketplace), each with its own regulatory profile. Encodes a repeatable end-to-end method from intake and regulatory mapping through framework architecture, ring-fencing, drafting, and signature across five phases and fourteen steps.
• cite-guardPaste an AI-drafted legal filing and cite-guard extracts every case citation, checks each against a known-cases index, and returns a single hallucination-risk score plus a per-citation verdict: verified, unverified, or fabricated. One number tells you whether a brief is safe to file. Open source, built to stop the fabricated citations courts have sanctioned lawyers for filing.
• Billable TimeWhen your bar comes asking "show me how you billed AI-assisted work" (and ABA 512, Florida 24-1, California, New York, and DC all have opinions out) you need an artifact that survives review. billable-time produces it. From your Claude Code session logs, it drafts reviewable time entries plus a printable HTML audit packet with: SHA-256 chain of evidence (source files + matter.yml + active disclosure pack + verifiable artifact self-hash), attorney identity and signature block, a bar-opinion disclosure pack with starter language for five jurisdictions, and content-aware deterministic narratives derived from filename and tool shape, never from prompt text by default. The tool refuses to bill on its own. --strict mode refuses to ship the artifact if any audit invariant fails (broad routes, missing attorney, missing/unverified disclosure). Comes as a Node CLI and a self-contained browser version (no backend; JSONL never leaves the page). 15 invariant tests verify the contract. AGPL-3.0.
• IRAC Prompt BuilderRestructure any rough build, research, or legal-drafting request into an IRAC-shaped prompt (Issue, Rule, Analysis, Conclusion) optimized for a frontier model. It's the bar-exam framework, repurposed as prompt engineering. The skill leads with the issue and ends with the conclusion (where models weight attention most), forces you to name your constraints and non-goals, and specifies what "good" looks like before a single token is generated. Use it before any non-trivial build, or whenever a vague ask deserves a precise brief.
• Oral ArgumentTribunal-prep skill modeled on Neal Katyal's Nov 2025 SCOTUS tariffs argument and his AI sparring partner "Harvey." Five phases: (1) profile each decision-maker from prior opinions, questions, concurrences, and dissents, surface their doctrinal commitments and institutional concerns; (2) predict the bench, 3–7 likely questions per judge with the worry behind each, verbatim phrasing where confidence is high; (3) map escape routes, the narrowest holding each skeptical judge could sign without abandoning prior commitments; (4) spar adversarially, relentless follow-ups, calls out parroted lines, tracks which questions broke the lawyer; (5) hand the lawyer back to themselves with a < 150-word index card for the podium. The model is the sparring partner. The lawyer still wins the case. Use for oral argument, motion hearings, depositions, arbitrations, or stress-testing a brief before filing.

Skills that do related work

• Legal AI Model RouterEntry point for routing any legal task to the right LLM, vendor-neutral and grounded in mid-2026 legal benchmarks. Works out which of five legal verticals a tas
• Route: Contract DraftingPick the right LLM for contract drafting: generating, redlining or rewriting contract language from instructions. Vendor-neutral routing grounded in mid-2026 le
• Route: Contract ReviewPick the right LLM for contract review: assessing an existing agreement for risk, deviations from standard, missing protections and internal contradictions, and
• Route: Info ExtractionPick the right LLM for legal info extraction: pulling facts, clauses, dates, parties, obligations and structured fields out of contracts and legal documents. Gr
• Route: CouncilEscalate a high-stakes legal task from a single model to a council: convene the top two or three, run them independently, score their agreement and synthesize a
• Route: Legal TranslationPick the right LLM for legal translation: contracts, statutes, case law and legal correspondence across languages, including Arabic and the wider MENA region. T