Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build a retrieval-informed chapter skeleton (`outline/chapter_skeleton.yml`) from taxonomy/core scope before stable H3 decomposition. **Trigger**: chapter skeleton, chapter-level outline, H2 skeleton, section-first survey, 章节骨架, 章级骨架. **Use when**: survey structure should stabilize chapter-level intent before subsection mapping and writing cards.
.claude/skills/willoscar-chapter-skeleton/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -67% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -19% | 0% |
Create outline/chapter_skeleton.refined.ok only after reviewing a manually refined skeleton. A changed taxonomy, goal, or generator invalidates the marker; reruns then back up and rebuild the skeleton.
Always read:
references/overview.mdUse scripts/run.py only for deterministic materialization:
outline/taxonomy.yml for retrieval-informed topic structureGOAL.md when present for scope hintsoutline/chapter_skeleton.ymloutline/taxonomy.ymlGOAL.mdoutline/chapter_skeleton.ymlassets/output_contract.jsonuv run python .codex/skills/chapter-skeleton/scripts/run.py --workspace <workspace>--workspace <dir>: workspace containing outline/taxonomy.yml--unit-id <id>: optional harness metadata--inputs <semicolon-separated>: optional override from UNITS.csv--outputs <semicolon-separated>: optional output override; default is outline/chapter_skeleton.yml--checkpoint <C*>: optional harness metadatauv run python .codex/skills/chapter-skeleton/scripts/run.py --workspace <workspace> --inputs 'outline/taxonomy.yml;GOAL.md' --outputs 'outline/chapter_skeleton.yml'| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 19,863 | 6,003 | -70% | 1 | 1 | 0% | 3,747 | 714 | -81% | 0 | 0 | — |
case-02 | fail→fail | 6,800 | 5,147 | -24% | 1 | 1 | 0% | 912 | 681 | -25% | 0 | 0 | — |
case-03 | fail→fail | 13,127 | 5,343 | -59% | 1 | 1 | 0% | 2,290 | 653 | -71% | 0 | 0 | — |
case-04 | fail→pass | 7,368 | 4,165 | -43% | 1 | 1 | 0% | 1,328 | 920 | -31% | 0 | 0 | — |
case-05 | pass→pass | 7,543 | 1,674 | -78% | 1 | 1 | 0% | 1,182 | 632 | -47% | 0 | 0 | — |
case-06 | pass→pass | 8,794 | 2,052 | -77% | 1 | 1 | 0% | 1,427 | 692 | -52% | 0 | 0 | — |
case-07 | pass→pass | 9,394 | 1,711 | -82% | 1 | 1 | 0% | 1,396 | 621 | -56% | 0 | 0 | — |
case-08 | fail→pass | 11,604 | 1,382 | -88% | 1 | 1 | 0% | 1,665 | 549 | -67% | 0 | 0 | — |
case-09 | pass→pass | 8,362 | 2,812 | -66% | 1 | 1 | 0% | 1,256 | 852 | -32% | 0 | 0 | — |
case-10 | pass→pass | 9,841 | 3,441 | -65% | 1 | 1 | 0% | 1,482 | 969 | -35% | 0 | 0 | — |
case-11 | fail→pass | 8,572 | 1,529 | -82% | 1 | 1 | 0% | 1,351 | 560 | -59% | 0 | 0 | — |
case-12 | fail→pass | 8,609 | 2,708 | -69% | 1 | 1 | 0% | 1,304 | 775 | -41% | 0 | 0 | — |
case-13 | pass→pass | 4,815 | 3,356 | -30% | 1 | 1 | 0% | 753 | 909 | +21% | 0 | 0 | — |
case-14 | fail→pass | 5,663 | 2,066 | -64% | 1 | 1 | 0% | 767 | 625 | -19% | 0 | 0 | — |
case-15 | pass→pass | 10,713 | 1,932 | -82% | 1 | 1 | 0% | 1,654 | 641 | -61% | 0 | 0 | — |
case-16 | fail→pass | 7,664 | 2,060 | -73% | 1 | 1 | 0% | 1,125 | 655 | -42% | 0 | 0 | — |
case-17 | pass→pass | 10,455 | 2,660 | -75% | 1 | 1 | 0% | 1,655 | 804 | -51% | 0 | 0 | — |
case-22 | fail→fail | 3,290 | 3,664 | +11% | 1 | 1 | 0% | 536 | 947 | +77% | 0 | 0 | — |
case-18 | fail→pass | 12,629 | 5,467 | -57% | 1 | 1 | 0% | 1,853 | 1,206 | -35% | 0 | 0 | — |
case-19 | fail→fail | 7,936 | 2,851 | -64% | 1 | 1 | 0% | 1,251 | 792 | -37% | 0 | 0 | — |
case-20 | pass→pass | 3,192 | 3,336 | +5% | 1 | 1 | 0% | 464 | 923 | +99% | 0 | 0 | — |
case-21 | pass→pass | 4,103 | 2,976 | -27% | 1 | 1 | 0% | 670 | 816 | +22% | 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 +32 percentage points is the difference between those two pass rates over the 19 comparable cases.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.