Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Analyze counterparty contract changes against agreement/schedules, flag conflicts, and produce severity-rated counterpositions.
.claude/skills/cowork-os-legal-contract-negotiation-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 983% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 919% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -12% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 4% | 0% |
Analyze counterparty contract changes against agreement/schedules, flag conflicts, and produce severity-rated counterpositions.
| Name | Type | Required | Description | |---|---|---|---| | agreement_path | string | No | Optional explicit path to the main agreement (DOCX/PDF/TXT) | | disclosure_schedules_path | string | No | Optional explicit path to disclosure schedules or side letter | | counterparty_changes_path | string | No | Optional explicit path to counterparty redline or demand letter | | client_side | select | No | Which side you represent | | output_report_path | string | No | Where to write the negotiation analysis markdown | | output_docx_path | string | No | Optional path for a clean revised DOCX draft |
../legal-contract-negotiation-review.json.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 6,307 | 16,588 | +163% | 1 | 1 | 0% | 313 | 3,389 | +983% | 0 | 0 | — |
case-02 | fail→pass | 6,679 | 23,943 | +258% | 1 | 1 | 0% | 442 | 4,503 | +919% | 0 | 0 | — |
case-03 | pass→pass | 16,242 | 11,901 | -27% | 1 | 1 | 0% | 2,906 | 2,567 | -12% | 0 | 0 | — |
case-04 | pass→pass | 18,402 | 17,453 | -5% | 1 | 1 | 0% | 3,444 | 3,575 | +4% | 0 | 0 | — |
case-05 | pass→pass | 26,982 | 17,979 | -33% | 1 | 1 | 0% | 4,529 | 3,531 | -22% | 0 | 0 | — |
case-06 | pass→pass | 16,266 | 14,566 | -10% | 1 | 1 | 0% | 2,582 | 2,743 | +6% | 0 | 0 | — |
case-07 | pass→pass | 14,048 | 21,242 | +51% | 1 | 1 | 0% | 2,156 | 3,861 | +79% | 0 | 0 | — |
case-08 | fail→pass | 14,042 | 20,384 | +45% | 1 | 1 | 0% | 2,243 | 3,800 | +69% | 0 | 0 | — |
case-09 | pass→pass | 16,741 | 18,504 | +11% | 1 | 1 | 0% | 2,614 | 3,327 | +27% | 0 | 0 | — |
case-10 | pass→pass | 14,505 | 20,472 | +41% | 1 | 1 | 0% | 2,146 | 3,695 | +72% | 0 | 0 | — |
case-11 | pass→pass | 12,193 | 15,665 | +28% | 1 | 1 | 0% | 1,931 | 2,975 | +54% | 0 | 0 | — |
case-12 | fail→fail | 12,590 | 16,264 | +29% | 1 | 1 | 0% | 2,096 | 3,048 | +45% | 0 | 0 | — |
case-13 | pass→pass | 14,451 | 20,241 | +40% | 1 | 1 | 0% | 2,393 | 3,690 | +54% | 0 | 0 | — |
case-14 | pass→pass | 15,692 | 18,150 | +16% | 1 | 1 | 0% | 2,374 | 3,134 | +32% | 0 | 0 | — |
case-20 | pass→pass | 15,999 | 18,255 | +14% | 1 | 1 | 0% | 2,475 | 3,183 | +29% | 0 | 0 | — |
case-15 | pass→pass | 12,700 | 19,880 | +57% | 1 | 1 | 0% | 1,923 | 3,400 | +77% | 0 | 0 | — |
case-16 | pass→pass | 12,723 | 13,364 | +5% | 1 | 1 | 0% | 1,899 | 2,702 | +42% | 0 | 0 | — |
case-17 | pass→pass | 14,275 | 16,747 | +17% | 1 | 1 | 0% | 2,153 | 2,869 | +33% | 0 | 0 | — |
case-18 | pass→pass | 13,691 | 26,072 | +90% | 1 | 1 | 0% | 2,134 | 4,419 | +107% | 0 | 0 | — |
case-19 | pass→pass | 14,427 | 14,906 | +3% | 1 | 1 | 0% | 2,213 | 2,661 | +20% | 0 | 0 | — |
case-21 | pass→pass | 13,006 | 15,587 | +20% | 1 | 1 | 0% | 1,895 | 2,902 | +53% | 0 | 0 | — |
case-22 | pass→pass | 15,951 | 19,464 | +22% | 1 | 1 | 0% | 2,459 | 3,228 | +31% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +14 percentage points is the difference between those two pass rates over the 20 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.