Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Turn an RFP and a folder of your past proposals, case studies, and capability docs into a requirement-by-requirement compliance matrix and a drafted response. Flags disqualifiers and unanswerable requirements before you burn a week writing.
.claude/skills/onewave-ai-cowork-rfp-response/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 1036% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 113% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 138% | 0% |
Respond to an RFP the way a seasoned proposal manager does: decompose every requirement first, decide bid/no-bid honestly, reuse the best prior language, and never claim a capability the source documents do not support. Inputs: the RFP document and a folder of company material (past proposals, case studies, bios, certifications, pricing sheets).
STRONG (direct evidence), PARTIAL (adjacent evidence, needs framing), or GAP (nothing found).STRONG items, adapt the best prior language to this client's context. For PARTIAL, draft honest framing and flag it for review. For GAP, insert a clearly marked [GAP: needs input -- suggested approach] block rather than fiction.compliance-matrix.md (the full matrix with response locations) and rfp-response-draft.md.GAP blocks are the honest alternative.questions-to-submit.md list, drafted in submission-ready form.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-18 | fail→fail | 13,601 | 6,416 | -53% | 1 | 1 | 0% | 2,207 | 1,648 | -25% | 0 | 0 | — |
case-01 | fail→pass | 5,122 | 20,002 | +291% | 1 | 1 | 0% | 284 | 3,226 | +1036% | 0 | 0 | — |
case-02 | fail→pass | 44,440 | 33,103 | -26% | 1 | 1 | 0% | 6,224 | 6,081 | -2% | 0 | 0 | — |
case-03 | pass→pass | 16,275 | 15,299 | -6% | 1 | 1 | 0% | 2,542 | 2,848 | +12% | 0 | 0 | — |
case-04 | fail→fail | 6,606 | 7,046 | +7% | 1 | 1 | 0% | 1,071 | 1,752 | +64% | 0 | 0 | — |
case-05 | fail→fail | 4,867 | 6,800 | +40% | 1 | 1 | 0% | 699 | 1,683 | +141% | 0 | 0 | — |
case-06 | fail→fail | 5,741 | 3,741 | -35% | 1 | 1 | 0% | 917 | 1,258 | +37% | 0 | 0 | — |
case-07 | fail→pass | 14,829 | 25,478 | +72% | 1 | 1 | 0% | 2,325 | 4,958 | +113% | 0 | 0 | — |
case-08 | fail→pass | 15,242 | 10,903 | -28% | 1 | 1 | 0% | 2,105 | 2,243 | +7% | 0 | 0 | — |
case-09 | pass→pass | 13,105 | 14,333 | +9% | 1 | 1 | 0% | 1,819 | 2,776 | +53% | 0 | 0 | — |
case-10 | fail→fail | 3,359 | 7,741 | +130% | 1 | 1 | 0% | 523 | 1,778 | +240% | 0 | 0 | — |
case-11 | pass→pass | 40,028 | 21,500 | -46% | 1 | 1 | 0% | 6,177 | 4,089 | -34% | 0 | 0 | — |
case-12 | pass→pass | 16,638 | 12,310 | -26% | 1 | 1 | 0% | 2,287 | 2,358 | +3% | 0 | 0 | — |
case-13 | fail→fail | 3,570 | 9,320 | +161% | 1 | 1 | 0% | 501 | 1,913 | +282% | 0 | 0 | — |
case-14 | fail→pass | 35,833 | 17,863 | -50% | 1 | 1 | 0% | 1,444 | 3,434 | +138% | 0 | 0 | — |
case-15 | fail→fail | 4,080 | 6,772 | +66% | 1 | 1 | 0% | 636 | 1,652 | +160% | 0 | 0 | — |
case-16 | pass→pass | 14,481 | 10,573 | -27% | 1 | 1 | 0% | 2,228 | 2,353 | +6% | 0 | 0 | — |
case-17 | fail→pass | 13,944 | 11,119 | -20% | 1 | 1 | 0% | 1,942 | 2,204 | +13% | 0 | 0 | — |
case-19 | pass→pass | 17,218 | 20,672 | +20% | 1 | 1 | 0% | 3,051 | 4,072 | +33% | 0 | 0 | — |
case-20 | fail→pass | 36,249 | 31,354 | -14% | 1 | 1 | 0% | 5,706 | 5,842 | +2% | 0 | 0 | — |
case-21 | pass→pass | 10,806 | 12,589 | +17% | 1 | 1 | 0% | 1,683 | 2,638 | +57% | 0 | 0 | — |
case-22 | pass→pass | 12,943 | 14,227 | +10% | 1 | 1 | 0% | 1,980 | 2,801 | +41% | 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 21 counted toward the lift figure. The other 1 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 +32 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.