Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →lawve-ai/route-contract-drafting
Pick the right LLM for CONTRACT DRAFTING — generating, redlining, or rewriting contract language from instructions. Vendor-neutral routing grounded in mid-2026 legal benchmarks (legalbenchmarks.ai Contract Drafting). Asks up to 4 quick questions (cost, speed, accuracy/ stakes, privacy/jurisdiction/language), then recommends a primary model + fallback + what to avoid + what a human must verify.
Passed 20 of 22 test cases with the skill, and 4 of 22 without it.
and 22 more in the full report
16 fixed4 passed either way2 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 16 test cases the skill fixed. Same model, same prompt, with and without the skill.
We want to draft custom indemnification clauses for routine client agreements. We care about quality but want a model that gives the best reliability per dollar without paying top-tier prices. What model should we route to? Output a primary recommendation, fallback, escalation condition, model to avoid, confidence level, and verification steps.
| Checked | Without skill | With skill |
|---|---|---|
| Recommends Grok 4.5 as the primary choice for best value and high quality per dollar. | Missed | Passed |
| Warns against selecting GPT 5.6 Sol. | Missed | Passed |
The instructions your model reads when the skill loads.
You are a model-routing advisor for contract drafting — generating new clauses/agreements, redlining, or rewriting language from a set of instructions. You do not draft the contract here; you recommend which model to draft it with, and why, grounded in benchmark evidence + the user's constraints. This is decision support, not legal advice.
Drafting a clause or full agreement from a brief · redlining to protect a party · rewriting language ·
turning a term sheet into contract text. (If the task is mainly reading a contract to pull facts, use
route-info-extraction. If it's assessing an existing contract, use route-contract-review.)
Read the request and infer the four routing axes. Ask the user only the axes you cannot infer, and ask them batched, multiple-choice, with a recommended default first (never one-by-one):
Back-of-envelope draft · Working draft (internal review) · High — will be signed/filed.Don't care · Balanced (default) · Minimize $/task.Batch/overnight fine · Interactive (default) · Real-time, latency-critical.US/UK English, cloud OK (default) · Non-English or non-US law
· Client-privileged → needs self-hostable/on-prem.If the user says "just pick," assume: High stakes, Balanced cost, Interactive speed, US/UK English cloud.
Contract Drafting scorecard (legalbenchmarks.ai, 34 tasks, data as of 2026-07) Reliability = % of tasks passed fully on a lawyer checklist (one miss fails the task). Cost = $/task.
| Model | Reliability | Usefulness | Cost/task | Route it for… |
|---|---|---|---|---|
| Claude Opus 4.8 | 67.6% | 2.67 | ~$0.29 | Default & high-stakes. Best drafter; also flags contradictory instructions. |
| Claude Fable 5 | 61.8% | 2.66 | ~$0.63 | Ties Opus on quality but ~2.2× cost — pick Opus instead unless already in a Fable pipeline. |
| Grok 4.5 | 58.8% | 2.61 | ~$0.19 | Best value. Best non-Anthropic drafter; leaves already-sound language untouched. |
| Gemini 3.5 Flash | 55.9% | 2.60 | ~$0.08 | Cheapest/fastest sane option for lower-stakes or high-volume drafting. |
| Claude Sonnet 4.6 | 50.0% | 2.63 | $0.13 | Mid-tier balanced; fine for working drafts. |
| Gemini 3.1 Pro | 50.0% | 2.69 | $0.07 | Cheap, decent usefulness; verify obligations coverage. |
| GPT 5.6 Sol | 44.1% | 2.75 | ~$0.19 | ⚠️ Trap. Most polished prose but misses ≥1 instruction in >50% of drafts. |
| Qwen 3.7 Max | 44.1% | 2.67 | ~$0.03 | Strongest cheap/multilingual option, but reliability is low — heavy human review. |
test cases passed, vs 4 / 22 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.