Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when creating or developing anything, before writing code or implementation plans - refines rough ideas into fully-formed designs through structured Socratic questioning, alternative exploration, and incremental validation
.claude/skills/aiskillstore-brainstorming/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -19% | 0% |
Turn raw ideas into clear, validated designs and specifications through structured dialogue before any implementation begins.
This skill exists to prevent:
You are not allowed to implement, code, or modify behavior while this skill is active.
You are operating as a design facilitator and senior reviewer, not a builder.
Your job is to slow the process down just enough to get it right.
Before asking any questions:
Do not design yet.
Your goal here is shared clarity, not speed.
Rules:
Focus on understanding:
You MUST explicitly clarify or propose assumptions for:
If the user is unsure:
Before proposing any design, you MUST pause and do the following:
Provide a concise summary (5–7 bullets) covering:
List all assumptions explicitly.
List unresolved questions, if any.
Then ask:
> “Does this accurately reflect your intent? > Please confirm or correct anything before we move to design.”
Do NOT proceed until explicit confirmation is given.
Once understanding is confirmed:
This is still not final design.
When presenting the design:
> “Does this look right so far?”
Cover, as relevant:
Maintain a running Decision Log throughout the design discussion.
For each decision:
This log should be preserved for documentation.
Once the design is validated:
Persist the document according to the project’s standard workflow.
Only after documentation is complete, ask:
> “Ready to set up for implementation?”
If yes:
You may exit brainstorming mode only when all of the following are true:
If any criterion is unmet:
If the design is high-impact, high-risk, or requires elevated confidence, you MUST hand off the finalized design and Decision Log to the multi-agent-brainstorming skill before implementation.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→pass | 15,917 | 4,962 | -69% | 1 | 1 | 0% | 1,690 | 2,026 | +20% | 0 | 0 | — |
case-19 | pass→pass | 17,760 | 10,992 | -38% | 1 | 1 | 0% | 1,889 | 1,969 | +4% | 0 | 0 | — |
case-01 | fail→pass | 25,804 | 10,407 | -60% | 1 | 1 | 0% | 4,156 | 1,949 | -53% | 0 | 0 | — |
case-02 | fail→pass | 17,102 | 6,436 | -62% | 1 | 1 | 0% | 1,822 | 2,036 | +12% | 0 | 0 | — |
case-08 | pass→pass | 13,182 | 12,361 | -6% | 1 | 1 | 0% | 1,282 | 2,346 | +83% | 0 | 0 | — |
case-03 | fail→pass | 11,491 | 9,618 | -16% | 1 | 1 | 0% | 1,711 | 1,722 | +1% | 0 | 0 | — |
case-04 | pass→fail | 14,750 | 6,469 | -56% | 1 | 1 | 0% | 2,240 | 2,071 | -8% | 0 | 0 | — |
case-05 | fail→pass | 23,739 | 13,671 | -42% | 1 | 1 | 0% | 3,013 | 2,452 | -19% | 0 | 0 | — |
case-06 | fail→fail | 23,053 | 15,493 | -33% | 1 | 1 | 0% | 3,039 | 2,697 | -11% | 0 | 0 | — |
case-07 | fail→fail | 53,370 | 8,376 | -84% | 1 | 1 | 0% | 8,231 | 2,483 | -70% | 0 | 0 | — |
case-10 | pass→fail | 20,775 | 9,604 | -54% | 1 | 1 | 0% | 2,203 | 1,835 | -17% | 0 | 0 | — |
case-11 | fail→fail | 11,314 | 12,619 | +12% | 1 | 1 | 0% | 1,069 | 2,399 | +124% | 0 | 0 | — |
case-12 | fail→pass | 13,208 | 8,286 | -37% | 1 | 1 | 0% | 1,337 | 1,548 | +16% | 0 | 0 | — |
case-13 | fail→pass | 13,581 | 11,115 | -18% | 1 | 1 | 0% | 2,168 | 2,077 | -4% | 0 | 0 | — |
case-14 | pass→pass | 11,461 | 1,891 | -84% | 1 | 1 | 0% | 968 | 1,414 | +46% | 0 | 0 | — |
case-15 | pass→pass | 20,452 | 13,483 | -34% | 1 | 1 | 0% | 2,441 | 2,541 | +4% | 0 | 0 | — |
case-16 | fail→pass | 8,747 | 8,861 | +1% | 1 | 1 | 0% | 528 | 1,660 | +214% | 0 | 0 | — |
case-17 | pass→pass | 14,546 | 3,437 | -76% | 1 | 1 | 0% | 1,419 | 1,685 | +19% | 0 | 0 | — |
case-18 | pass→pass | 29,805 | 19,320 | -35% | 1 | 1 | 0% | 3,529 | 3,228 | -9% | 0 | 0 | — |
case-20 | fail→pass | 11,850 | 4,717 | -60% | 1 | 1 | 0% | 2,213 | 1,824 | -18% | 0 | 0 | — |
case-21 | pass→fail | 8,561 | 9,866 | +15% | 1 | 1 | 0% | 669 | 1,996 | +198% | 0 | 0 | — |
case-22 | pass→fail | 8,367 | 13,693 | +64% | 1 | 1 | 0% | 737 | 2,592 | +252% | 0 | 0 | — |
case-23 | pass→fail | 7,994 | 13,798 | +73% | 1 | 1 | 0% | 495 | 2,248 | +354% | 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. 23 cases were attempted. The headline lift of +17 percentage points is the difference between those two pass rates over the 23 comparable cases. 5 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.