Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Screen the direction pool with a discussion-first scoring pass, writing `output/trace/IDEA_SCREENING_TABLE.md`. **Trigger**: idea screener, screening table, brainstorm screening, 方向筛选表. **Use when**: you already have a direction pool and want a table-first comparison before curating the shortlist.
.claude/skills/willoscar-idea-screener/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -45% | 0% |
Goal: compress a direction pool into a scored comparison table that helps shortlist the most discussion-worthy directions.
The screener should reward:
and penalize same-template directions that only swap nouns.
uv run python .codex/skills/idea-screener/scripts/run.py --workspace <workspace>--workspace <dir> (required)--unit-id <U###>--inputs <semicolon-separated>--outputs <semicolon-separated>--checkpoint <C#>uv run python .codex/skills/idea-screener/scripts/run.py --workspace workspaces/brainstorm-llm-agents| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | pass→pass | 18,539 | 18,175 | -2% | 1 | 1 | 0% | 2,593 | 2,647 | +2% | 0 | 0 | — |
case-01 | fail→fail | 6,550 | 4,715 | -28% | 1 | 1 | 0% | 961 | 463 | -52% | 0 | 0 | — |
case-02 | fail→fail | 6,985 | 4,491 | -36% | 1 | 1 | 0% | 1,002 | 389 | -61% | 0 | 0 | — |
case-03 | fail→fail | 6,675 | 3,630 | -46% | 1 | 1 | 0% | 977 | 330 | -66% | 0 | 0 | — |
case-04 | pass→pass | 7,907 | 1,712 | -78% | 1 | 1 | 0% | 1,246 | 471 | -62% | 0 | 0 | — |
case-05 | pass→pass | 8,940 | 2,330 | -74% | 1 | 1 | 0% | 1,481 | 576 | -61% | 0 | 0 | — |
case-06 | fail→pass | 6,922 | 1,767 | -74% | 1 | 1 | 0% | 1,177 | 469 | -60% | 0 | 0 | — |
case-07 | fail→pass | 7,303 | 2,929 | -60% | 1 | 1 | 0% | 1,283 | 704 | -45% | 0 | 0 | — |
case-08 | fail→pass | 7,839 | 1,843 | -76% | 1 | 1 | 0% | 1,292 | 468 | -64% | 0 | 0 | — |
case-09 | fail→pass | 4,226 | 2,194 | -48% | 1 | 1 | 0% | 639 | 520 | -19% | 0 | 0 | — |
case-10 | pass→pass | 21,065 | 21,972 | +4% | 1 | 1 | 0% | 2,972 | 3,422 | +15% | 0 | 0 | — |
case-11 | pass→pass | 18,970 | 17,389 | -8% | 1 | 1 | 0% | 2,647 | 2,645 | -0% | 0 | 0 | — |
case-12 | pass→pass | 20,091 | 11,443 | -43% | 1 | 1 | 0% | 3,580 | 2,197 | -39% | 0 | 0 | — |
case-14 | pass→pass | 13,778 | 13,503 | -2% | 1 | 1 | 0% | 1,922 | 2,025 | +5% | 0 | 0 | — |
case-15 | fail→fail | 12,044 | 5,049 | -58% | 1 | 1 | 0% | 1,621 | 867 | -47% | 0 | 0 | — |
case-16 | pass→pass | 15,358 | 14,644 | -5% | 1 | 1 | 0% | 2,110 | 2,195 | +4% | 0 | 0 | — |
case-17 | fail→pass | 6,850 | 2,763 | -60% | 1 | 1 | 0% | 1,125 | 624 | -45% | 0 | 0 | — |
case-18 | fail→pass | 6,833 | 2,067 | -70% | 1 | 1 | 0% | 1,054 | 535 | -49% | 0 | 0 | — |
case-19 | pass→pass | 7,380 | 2,446 | -67% | 1 | 1 | 0% | 1,307 | 631 | -52% | 0 | 0 | — |
case-20 | fail→fail | 7,437 | 2,678 | -64% | 1 | 1 | 0% | 1,042 | 574 | -45% | 0 | 0 | — |
case-21 | fail→pass | 6,413 | 3,500 | -45% | 1 | 1 | 0% | 959 | 672 | -30% | 0 | 0 | — |
case-22 | fail→pass | 7,565 | 1,718 | -77% | 1 | 1 | 0% | 1,207 | 406 | -66% | 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 19 counted toward the lift figure. The other 3 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 +36 percentage points is the difference between those two pass rates over the 19 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.