Install any skill in seconds. Free to start, no credit card required.
Get Started Free →- Produce a complete long-form fiction pipeline with minimal back-and-forth.
.claude/skills/cowork-os-novelist/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 90% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 252% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-07 | ✓→✗ | ▼ Worse | 4% | 0% |
seedgenre, target_words, chapter_count, packaging, canon_modeIf the user invoked /novelist without optional parameters, use the defaults and continue. If seed is missing, ask only for the seed and pause.
artifacts/skills/novelist/Expected files:
brief.mdworld.mdcharacters.mdoutline.mdvoice.mdcanon.mdprogress.jsonchapters/chapter-XX-title.mdrevision-notes.mdmanuscript_complete.mdmanifest.md.epub and .pdfbrief.md.fanfiction or inspired-by.world.md: setting, factions, technology/magic rules, political pressures.characters.md: core cast, motives, conflicts, arcs.voice.md: prose rules, pacing, dialogue texture, taboo phrasing to avoid.canon.md: continuity constraints, timeline anchors, terminology.outline.md: chapter-by-chapter plan with purpose and cliffhangers.progress.json with chapter count, estimated word count, and current phase.manuscript_complete.md from the chapter files.revision-notes.md with continuity fixes, pacing fixes, and weak spots addressed.generate_epub for ebook export.generate_document for PDF only after manuscript_complete.md is ready.manifest.md.manuscript_complete.md exists.manifest.md exists and lists files produced plus remaining blockers.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 46,535 | 45,838 | -1% | 1 | 1 | 0% | 6,215 | 7,107 | +14% | 0 | 0 | — |
case-02 | fail→fail | 47,014 | 70,891 | +51% | 1 | 1 | 0% | 6,204 | 1,597 | -74% | 0 | 0 | — |
case-03 | fail→fail | 50,097 | 12,312 | -75% | 1 | 1 | 0% | 6,223 | 1,621 | -74% | 0 | 0 | — |
case-04 | pass→pass | 7,476 | 3,132 | -58% | 1 | 1 | 0% | 1,188 | 1,305 | +10% | 0 | 0 | — |
case-05 | fail→pass | 12,172 | 11,156 | -8% | 1 | 1 | 0% | 1,314 | 2,502 | +90% | 0 | 0 | — |
case-06 | fail→fail | 23,799 | 7,744 | -67% | 1 | 1 | 0% | 3,309 | 2,099 | -37% | 0 | 0 | — |
case-07 | pass→fail | 9,285 | 6,633 | -29% | 1 | 1 | 0% | 1,447 | 1,509 | +4% | 0 | 0 | — |
case-08 | fail→pass | 14,592 | 44,539 | +205% | 1 | 1 | 0% | 2,011 | 7,077 | +252% | 0 | 0 | — |
case-09 | fail→pass | 11,920 | 8,862 | -26% | 1 | 1 | 0% | 1,832 | 2,306 | +26% | 0 | 0 | — |
case-10 | pass→pass | 6,080 | 8,488 | +40% | 1 | 1 | 0% | 843 | 2,217 | +163% | 0 | 0 | — |
case-11 | pass→pass | 11,801 | 2,128 | -82% | 1 | 1 | 0% | 1,640 | 1,207 | -26% | 0 | 0 | — |
case-12 | fail→fail | 7,129 | 2,791 | -61% | 1 | 1 | 0% | 1,025 | 1,339 | +31% | 0 | 0 | — |
case-13 | fail→pass | 9,284 | 2,960 | -68% | 1 | 1 | 0% | 1,444 | 1,347 | -7% | 0 | 0 | — |
case-14 | fail→fail | 9,542 | 2,730 | -71% | 1 | 1 | 0% | 1,624 | 1,316 | -19% | 0 | 0 | — |
case-15 | pass→pass | 5,564 | 2,430 | -56% | 1 | 1 | 0% | 887 | 1,267 | +43% | 0 | 0 | — |
case-16 | fail→fail | 10,926 | 3,386 | -69% | 1 | 1 | 0% | 1,718 | 1,535 | -11% | 0 | 0 | — |
case-17 | pass→pass | 18,460 | 13,305 | -28% | 1 | 1 | 0% | 2,720 | 2,810 | +3% | 0 | 0 | — |
case-18 | fail→fail | 9,537 | 3,006 | -68% | 1 | 1 | 0% | 1,401 | 1,380 | -1% | 0 | 0 | — |
case-19 | pass→pass | 12,665 | 4,750 | -62% | 1 | 1 | 0% | 1,906 | 1,640 | -14% | 0 | 0 | — |
case-20 | pass→pass | 3,889 | 14,923 | +284% | 1 | 1 | 0% | 688 | 3,346 | +386% | 0 | 0 | — |
case-21 | pass→pass | 8,409 | 6,412 | -24% | 1 | 1 | 0% | 1,396 | 1,864 | +34% | 0 | 0 | — |
case-22 | fail→fail | 3,752 | 4,690 | +25% | 1 | 1 | 0% | 576 | 1,683 | +192% | 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 +14 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is 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.