Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Run STORM phase 3 — per-section article writing. This skill should be used when the user asks to "write the article sections", "draft the storm article", or invokes /storm:write. Writes each outline section in parallel with inline citations grounded in the research sources.
.claude/skills/fradser-storm-write/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -84% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-06 | ✓→✗ | ▼ Worse | 19% | 0% |
Phase 3 of the STORM pipeline. Writes each outline section in parallel (one Task subagent per section), each grounded in top-k relevant sources from the research information table, with inline [n] citations.
storm-engine via the Skill tool.outline.md MUST exist (phase 2 complete). If absent, stop and instruct the user to run /storm:outline first.This phase is complete iff article.md exists and every outline section (except Introduction/Conclusion/Summary placeholders) has body text. If --force is not set and it exists, skip and exit early.
outline.md and research/sources.json. Index sources for retrieval (simple: rank by keyword/heading overlap with the section title).polish).--retrieve-top-k 3), the full outline for context.[n] citations mapping to sources.json ids.<!-- TODO: no source -->.article.md. Preserve the heading hierarchy from outline.md.run-config.json: phases.write = "completed", section count, any TODO-flagged sections.[n] where n is an existing id in sources.json. Never mint a new id here.[n] immediately after the claim it supports. Multi-source claims use [1][2].Report: number of sections written, number of citations used, list of sections flagged as TODO (no source), and the path to article.md.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 21,354 | 4,289 | -80% | 1 | 1 | 0% | 3,651 | 745 | -80% | 0 | 0 | — |
case-02 | fail→fail | 7,239 | 4,818 | -33% | 1 | 1 | 0% | 1,169 | 756 | -35% | 0 | 0 | — |
case-03 | fail→fail | 51,267 | 4,366 | -91% | 1 | 1 | 0% | 8,239 | 834 | -90% | 0 | 0 | — |
case-04 | fail→pass | 51,734 | 4,986 | -90% | 1 | 1 | 0% | 8,234 | 1,330 | -84% | 0 | 0 | — |
case-05 | fail→fail | 2,379 | 5,612 | +136% | 1 | 1 | 0% | 314 | 857 | +173% | 0 | 0 | — |
case-06 | pass→fail | 4,037 | 6,101 | +51% | 1 | 1 | 0% | 647 | 768 | +19% | 0 | 0 | — |
case-12 | fail→pass | 16,411 | 13,443 | -18% | 1 | 1 | 0% | 2,411 | 1,739 | -28% | 0 | 0 | — |
case-07 | fail→fail | 31,096 | 5,807 | -81% | 1 | 1 | 0% | 5,474 | 700 | -87% | 0 | 0 | — |
case-08 | fail→fail | 9,229 | 3,844 | -58% | 1 | 1 | 0% | 1,487 | 695 | -53% | 0 | 0 | — |
case-09 | fail→fail | 5,426 | 5,881 | +8% | 1 | 1 | 0% | 260 | 819 | +215% | 0 | 0 | — |
case-10 | fail→pass | 11,002 | 2,723 | -75% | 1 | 1 | 0% | 1,677 | 930 | -45% | 0 | 0 | — |
case-11 | fail→fail | 13,977 | 7,149 | -49% | 1 | 1 | 0% | 2,068 | 895 | -57% | 0 | 0 | — |
case-13 | pass→pass | 9,587 | 15,683 | +64% | 1 | 1 | 0% | 1,540 | 2,781 | +81% | 0 | 0 | — |
case-14 | fail→fail | 9,079 | 4,021 | -56% | 1 | 1 | 0% | 1,262 | 710 | -44% | 0 | 0 | — |
case-15 | fail→pass | 9,150 | 2,205 | -76% | 1 | 1 | 0% | 1,716 | 871 | -49% | 0 | 0 | — |
case-16 | pass→fail | 5,350 | 8,210 | +53% | 1 | 1 | 0% | 979 | 876 | -11% | 0 | 0 | — |
case-22 | fail→fail | 7,023 | 6,650 | -5% | 1 | 1 | 0% | 322 | 943 | +193% | 0 | 0 | — |
case-17 | pass→pass | 7,895 | 5,661 | -28% | 1 | 1 | 0% | 1,272 | 1,374 | +8% | 0 | 0 | — |
case-18 | fail→fail | 11,723 | 5,686 | -51% | 1 | 1 | 0% | 2,157 | 863 | -60% | 0 | 0 | — |
case-19 | pass→pass | 12,393 | 6,321 | -49% | 1 | 1 | 0% | 2,196 | 1,428 | -35% | 0 | 0 | — |
case-20 | pass→fail | 30,688 | 4,898 | -84% | 1 | 1 | 0% | 5,703 | 800 | -86% | 0 | 0 | — |
case-21 | fail→fail | 44,797 | 4,334 | -90% | 1 | 1 | 0% | 8,234 | 857 | -90% | 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 7 counted toward the lift figure. The other 15 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 +5 percentage points is the difference between those two pass rates over the 7 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.