Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Lock an ideation run into a single-source-of-truth brainstorm brief (`output/trace/IDEA_BRIEF.md`) and a replayable multi-query plan (`queries.md`). **Trigger**: idea brief, ideation brief, research ideas, brainstorm, 找 idea, 选题, 点子, 找方向. **Use when**: the user wants research ideas and their input is long / multi-turn; you need to clarify topic + constraints before retrieval.
.claude/skills/willoscar-idea-brief/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -58% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -54% | 0% |
Turn a fuzzy ideation request into an auditable brainstorm contract for the later ideation stack.
This skill does not retrieve papers. It only locks topic, audience, constraints, exclusions, rubric, query buckets, and target artifact shape into:
output/trace/IDEA_BRIEF.mdqueries.mdDECISIONS.mdGOAL.mdDECISIONS.mdqueries.mdoutput/trace/IDEA_BRIEF.mdqueries.mdDECISIONS.mdThis skill starts from GOAL.md, then refines the topic/constraints into output/trace/IDEA_BRIEF.md, updates queries.md, and records blockers or approvals in DECISIONS.md.
Read references/overview.md before changing the package shape or the brief contract. assets/brief_contract.json is the machine-readable source for:
uv run python .codex/skills/idea-brief/scripts/run.py --workspace <workspace>--workspace <dir>--unit-id <id>--inputs <a;b;...>--outputs <a;b;...>--checkpoint <C*>uv run python .codex/skills/idea-brief/scripts/run.py --workspace <workspace>| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | fail→fail | 59,498 | 58,784 | -1% | 1 | 1 | 0% | 1,451 | 5,794 | +299% | 0 | 0 | — |
case-01 | fail→fail | 15,039 | 5,315 | -65% | 1 | 1 | 0% | 2,355 | 635 | -73% | 0 | 0 | — |
case-02 | fail→fail | 3,468 | 5,207 | +50% | 1 | 1 | 0% | 279 | 688 | +147% | 0 | 0 | — |
case-03 | fail→fail | 5,091 | 4,174 | -18% | 1 | 1 | 0% | 322 | 589 | +83% | 0 | 0 | — |
case-04 | pass→fail | 8,907 | 8,011 | -10% | 1 | 1 | 0% | 1,560 | 806 | -48% | 0 | 0 | — |
case-05 | pass→fail | 9,765 | 5,384 | -45% | 1 | 1 | 0% | 1,491 | 640 | -57% | 0 | 0 | — |
case-07 | fail→pass | 5,456 | 1,716 | -69% | 1 | 1 | 0% | 873 | 639 | -27% | 0 | 0 | — |
case-08 | fail→pass | 21,724 | 1,761 | -92% | 1 | 1 | 0% | 1,776 | 642 | -64% | 0 | 0 | — |
case-09 | fail→pass | 10,367 | 2,030 | -80% | 1 | 1 | 0% | 1,673 | 709 | -58% | 0 | 0 | — |
case-10 | fail→pass | 8,847 | 2,452 | -72% | 1 | 1 | 0% | 1,382 | 726 | -47% | 0 | 0 | — |
case-11 | fail→pass | 9,455 | 2,122 | -78% | 1 | 1 | 0% | 1,512 | 701 | -54% | 0 | 0 | — |
case-12 | fail→pass | 11,385 | 1,795 | -84% | 1 | 1 | 0% | 1,976 | 665 | -66% | 0 | 0 | — |
case-13 | pass→pass | 7,359 | 1,879 | -74% | 1 | 1 | 0% | 1,151 | 674 | -41% | 0 | 0 | — |
case-14 | fail→pass | 8,518 | 1,417 | -83% | 1 | 1 | 0% | 1,251 | 565 | -55% | 0 | 0 | — |
case-15 | fail→pass | 9,379 | 1,354 | -86% | 1 | 1 | 0% | 1,467 | 527 | -64% | 0 | 0 | — |
case-16 | fail→pass | 12,838 | 2,680 | -79% | 1 | 1 | 0% | 1,958 | 751 | -62% | 0 | 0 | — |
case-17 | fail→pass | 14,944 | 3,229 | -78% | 1 | 1 | 0% | 2,243 | 887 | -60% | 0 | 0 | — |
case-18 | pass→pass | 11,601 | 3,203 | -72% | 1 | 1 | 0% | 1,575 | 833 | -47% | 0 | 0 | — |
case-19 | fail→fail | 6,407 | 4,322 | -33% | 1 | 1 | 0% | 900 | 1,082 | +20% | 0 | 0 | — |
case-20 | pass→pass | 12,867 | 4,434 | -66% | 1 | 1 | 0% | 1,783 | 996 | -44% | 0 | 0 | — |
case-21 | fail→fail | 9,408 | 2,347 | -75% | 1 | 1 | 0% | 1,396 | 707 | -49% | 0 | 0 | — |
case-22 | fail→pass | 15,113 | 1,689 | -89% | 1 | 1 | 0% | 995 | 671 | -33% | 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 15 counted toward the lift figure. The other 7 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 +41 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.