Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →MUST use before brainstorming, visual companion, Pencil/Figma, or implementation when a user gives a rough product, feature, app, website, landing page, brand site, front-end design draft, or "do not develop yet" goal. Draft the first structured requirement contract and fixed workspace before design or implementation.
.claude/skills/hashgraph-online-requirement-discovery/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 133% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -14% | 0% |
This skill turns a rough user goal into the first product-development contract. The user only needs to provide a rough goal, corrections, and confirmations.
The agent must draft information architecture, draft interaction structure, draft state requirements, draft API/data assumptions, and draft acceptance boundaries. Do not ask the user to provide IA, interaction matrices, state matrices, API contracts, or acceptance criteria from scratch.
For one-line site or design requests, start here. Examples:
Do not offer a browser visual companion, open a local URL, start Pencil/Figma, or run a creative-only brainstorming flow before this skill has produced the fixed requirement workspace through generate_requirement_workspace. If brainstorming is useful, capture the idea as AI Assumptions inside the requirement discovery artifact, then ask at most three focused confirmation questions.
docs/full-stack-development/requirements/<feature-slug>/01-intake/requirement-discovery.mdgenerate_requirement_workspacereview_artifact_contractreview_stage_gatebrainstorming replace requirement discovery for product, website, landing page, brand site, or design-draft requests.generate_requirement_workspace.rough-admin-one-line must produce a usable draft.over-clarification must block when the agent asks too many questions without a draft.empty-requirement-discovery must block.review_artifact_contract passes for requirement discovery.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 42,864 | 44,083 | +3% | 1 | 1 | 0% | 7,246 | 5,103 | -30% | 0 | 0 | — |
case-02 | fail→fail | 53,002 | 35,850 | -32% | 1 | 1 | 0% | 6,854 | 3,693 | -46% | 0 | 0 | — |
case-03 | fail→fail | 35,263 | 12,600 | -64% | 1 | 1 | 0% | 4,074 | 1,009 | -75% | 0 | 0 | — |
case-04 | pass→pass | 8,073 | 12,232 | +52% | 1 | 1 | 0% | 1,293 | 2,402 | +86% | 0 | 0 | — |
case-05 | pass→fail | 26,875 | 55,846 | +108% | 1 | 1 | 0% | 2,937 | 2,121 | -28% | 0 | 0 | — |
case-06 | pass→pass | 13,187 | 8,048 | -39% | 1 | 1 | 0% | 1,628 | 2,034 | +25% | 0 | 0 | — |
case-07 | fail→fail | 21,715 | 15,048 | -31% | 1 | 1 | 0% | 2,339 | 1,947 | -17% | 0 | 0 | — |
case-08 | fail→pass | 10,583 | 25,653 | +142% | 1 | 1 | 0% | 1,719 | 4,003 | +133% | 0 | 0 | — |
case-09 | fail→fail | 19,857 | 35,419 | +78% | 1 | 1 | 0% | 4,168 | 5,425 | +30% | 0 | 0 | — |
case-10 | fail→fail | 19,037 | 15,554 | -18% | 1 | 1 | 0% | 3,173 | 1,057 | -67% | 0 | 0 | — |
case-11 | fail→fail | 35,681 | 24,637 | -31% | 1 | 1 | 0% | 6,896 | 1,083 | -84% | 0 | 0 | — |
case-12 | fail→fail | 11,506 | 16,905 | +47% | 1 | 1 | 0% | 1,686 | 1,170 | -31% | 0 | 0 | — |
case-13 | fail→pass | 7,811 | 3,001 | -62% | 1 | 1 | 0% | 1,226 | 1,074 | -12% | 0 | 0 | — |
case-14 | fail→fail | 7,031 | 20,795 | +196% | 1 | 1 | 0% | 1,289 | 1,123 | -13% | 0 | 0 | — |
case-15 | pass→pass | 16,795 | 32,373 | +93% | 1 | 1 | 0% | 2,097 | 3,177 | +52% | 0 | 0 | — |
case-16 | fail→pass | 25,817 | 24,262 | -6% | 1 | 1 | 0% | 4,134 | 4,689 | +13% | 0 | 0 | — |
case-17 | fail→pass | 16,407 | 3,037 | -81% | 1 | 1 | 0% | 1,791 | 1,024 | -43% | 0 | 0 | — |
case-18 | fail→fail | 15,575 | 11,879 | -24% | 1 | 1 | 0% | 1,754 | 1,102 | -37% | 0 | 0 | — |
case-19 | fail→fail | 25,121 | 7,905 | -69% | 1 | 1 | 0% | 3,455 | 935 | -73% | 0 | 0 | — |
case-20 | fail→fail | 16,126 | 31,766 | +97% | 1 | 1 | 0% | 2,629 | 5,888 | +124% | 0 | 0 | — |
case-21 | fail→pass | 9,936 | 6,335 | -36% | 1 | 1 | 0% | 1,547 | 1,338 | -14% | 0 | 0 | — |
case-22 | fail→fail | 9,512 | 4,466 | -53% | 1 | 1 | 0% | 1,537 | 1,170 | -24% | 0 | 0 | — |
case-23 | fail→pass | 7,227 | 2,262 | -69% | 1 | 1 | 0% | 1,210 | 860 | -29% | 0 | 0 | — |
case-24 | pass→pass | 20,756 | 15,939 | -23% | 1 | 1 | 0% | 2,801 | 3,427 | +22% | 0 | 0 | — |
case-25 | fail→fail | 37,206 | 8,632 | -77% | 1 | 1 | 0% | 8,220 | 1,160 | -86% | 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. 25 cases were attempted, and 15 counted toward the lift figure. The other 10 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 +20 percentage points is the difference between those two pass rates over the 15 comparable cases. 3 cases got worse with the skill loaded, and they are 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.