Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Theory-driven literature review for building arguments and frameworks. Flexible, subjective, and narrative-focused — selects evidence strategically to support a thesis. High web-research budget for blogs, opinion pieces, and industry perspectives. Use when the user is writing a position paper, survey introduction, or constructing a coherent narrative around a research theme.
.claude/skills/yogsoth-ai-narrative-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 47% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 142% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -10% | 0% |
Purpose: Flexible, subjective, theory-driven — build a theoretical argument supported by literature. Not objective coverage, but persuasive framing.
When to use: User is writing a position paper, survey introduction, or needs to construct a coherent narrative around a research theme.
| Base SOP | Target | ±10% Range | |----------|--------|------------| | web-search | 80 results | 72–88 | | web-research | 15 pages | 13–17 | | paper-overview | 50 papers | 45–55 | | paper-search | 40 papers | 36–44 | | paper-research | 20 papers | 18–22 |
Print this table before each major iteration decision:
| SOP | Target | Current | % Complete |
|----------------|--------|---------|------------|
| web-search | 80 | ??? | ???% |
| web-research | 15 | ??? | ???% |
| paper-overview | 50 | ??? | ???% |
| paper-search | 40 | ??? | ???% |
| paper-research | 20 | ??? | ???% |Do not exit the strategy until all rows reach ≥90%.
narrative-framing — define themes → build argument → select supporting evidenceImport (strict protocol execution):
web-search → web-browsing/skills/web-search/SKILL.mdweb-research → web-browsing/skills/web-research/SKILL.mdpaper-overview → literature-engine/skills/literature-overview/SKILL.mdpaper-search → literature-engine/skills/literature-search/SKILL.mdpaper-research → literature-engine/skills/literature-research/SKILL.mdSubagent (CC decides when to invoke):
categorize-papers — cluster papers by theme/method/timelinethematic-coding — identify recurring themes across papersgap-identification — find what the literature hasn't addressedsurvey-synthesis — produce final structured outputStructured Narrative containing:
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | narrative-framing | Theory-driven reading tactic — define a theoretical framework first, then guide reading to fill it with evidence. Five stages (theme identification, argument construction, evidence collection, counter-evidence, synthesis). The most intellectually demanding tactic. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | categorize-papers | Cluster papers by theme, method, or timeline. Produces natural groupings from a paper collection. Used by scoping-survey and narrative-review. | | knowledge-acquisition-gap-identification | Identify what the literature has NOT addressed — missing methods, untested combinations, unexplored applications, contradictions without resolution. Used by all strategies. | | knowledge-acquisition-paper-overview | Abstract-level paper scanning for broad coverage. Import of literature-engine/literature-overview skill. Abstract-level only — no methodology conclusions from abstracts. | | knowledge-acquisition-paper-research | Full-depth paper reading with raw text extraction. Import of literature-engine/literature-research skill. Must read fullText (true) — equations, hyperparameters, specific claims extracted. | | knowledge-acquisition-paper-search | AI-summarized paper reading for intermediate depth. Import of literature-engine/literature-search skill. Must call get_paper_content for every analyzed paper. | | knowledge-acquisition-web-research | Full-page web reading for non-academic perspectives — blogs, tech reports, product pages, industry analysis. Import of web-browsing/web-research skill. Must fetch full page via apify for every analyzed page. | | knowledge-acquisition-web-search | Quick web scanning for landscape understanding. Import of web-browsing/web-search skill. Snippets only — no conclusions from snippets alone. | | survey-synthesis | Final synthesis step — weave all gathered evidence (reading notes, extracted data, categorizations) into a coherent structured output appropriate to the strategy type. Used by all 5 strategies as the final step. | | thematic-coding | Identify recurring themes across papers using qualitative coding methodology. Produces a codebook with theme definitions, supporting evidence, and frequency counts. Used by narrative-review. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 41,644 | 49,665 | +19% | 1 | 1 | 0% | 5,693 | 8,364 | +47% | 0 | 0 | — |
case-07 | fail→pass | 51,246 | 38,187 | -25% | 1 | 1 | 0% | 4,997 | 6,363 | +27% | 0 | 0 | — |
case-02 | fail→fail | 51,553 | 9,120 | -82% | 1 | 1 | 0% | 6,991 | 2,259 | -68% | 0 | 0 | — |
case-03 | fail→fail | 39,068 | 28,088 | -28% | 1 | 1 | 0% | 5,244 | 2,408 | -54% | 0 | 0 | — |
case-04 | pass→fail | 44,915 | 31,465 | -30% | 1 | 1 | 0% | 5,415 | 2,614 | -52% | 0 | 0 | — |
case-05 | pass→pass | 32,408 | 61,205 | +89% | 1 | 1 | 0% | 4,963 | 9,453 | +90% | 0 | 0 | — |
case-06 | pass→pass | 39,106 | 47,308 | +21% | 1 | 1 | 0% | 3,181 | 7,129 | +124% | 0 | 0 | — |
case-08 | fail→fail | 78,236 | 35,125 | -55% | 1 | 1 | 0% | 2,408 | 3,335 | +38% | 0 | 0 | — |
case-09 | fail→pass | 50,780 | 29,912 | -41% | 1 | 1 | 0% | 2,145 | 5,186 | +142% | 0 | 0 | — |
case-10 | fail→pass | 43,606 | 32,240 | -26% | 1 | 1 | 0% | 2,188 | 1,702 | -22% | 0 | 0 | — |
case-11 | fail→pass | 16,367 | 32,493 | +99% | 1 | 1 | 0% | 2,119 | 1,908 | -10% | 0 | 0 | — |
case-12 | fail→pass | 15,426 | 8,797 | -43% | 1 | 1 | 0% | 2,291 | 1,782 | -22% | 0 | 0 | — |
case-13 | fail→pass | 20,378 | 16,740 | -18% | 1 | 1 | 0% | 2,459 | 3,113 | +27% | 0 | 0 | — |
case-14 | pass→pass | 20,398 | 15,221 | -25% | 1 | 1 | 0% | 2,485 | 3,866 | +56% | 0 | 0 | — |
case-15 | pass→pass | 15,894 | 30,616 | +93% | 1 | 1 | 0% | 1,731 | 5,678 | +228% | 0 | 0 | — |
case-16 | pass→pass | 12,261 | 13,983 | +14% | 1 | 1 | 0% | 1,668 | 2,852 | +71% | 0 | 0 | — |
case-17 | pass→pass | 22,492 | 41,637 | +85% | 1 | 1 | 0% | 2,630 | 7,033 | +167% | 0 | 0 | — |
case-18 | fail→pass | 44,894 | 30,291 | -33% | 1 | 1 | 0% | 1,092 | 5,500 | +404% | 0 | 0 | — |
case-19 | fail→pass | 18,973 | 40,261 | +112% | 1 | 1 | 0% | 2,317 | 7,421 | +220% | 0 | 0 | — |
case-20 | pass→pass | 21,156 | 11,024 | -48% | 1 | 1 | 0% | 2,592 | 2,085 | -20% | 0 | 0 | — |
case-21 | pass→pass | 20,325 | 42,375 | +108% | 1 | 1 | 0% | 2,401 | 8,328 | +247% | 0 | 0 | — |
case-22 | pass→pass | 19,579 | 37,679 | +92% | 1 | 1 | 0% | 2,119 | 6,235 | +194% | 0 | 0 | — |
case-23 | fail→pass | 16,386 | 8,855 | -46% | 1 | 1 | 0% | 1,815 | 1,860 | +2% | 0 | 0 | — |
case-24 | fail→pass | 16,379 | 12,552 | -23% | 1 | 1 | 0% | 1,915 | 1,666 | -13% | 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. 24 cases were attempted, and 23 counted toward the lift figure. The other 1 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 +42 percentage points is the difference between those two pass rates over the 23 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.