Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Precise, targeted investigation of a specific sub-problem — few papers, all read in full depth. High paper-research ratio (50% deep-read rate). Use when the user knows exactly what they need to understand and requires detailed technical analysis with equations, hyperparameters, and specific claims extracted.
.claude/skills/yogsoth-ai-deep-survey/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 104% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -70% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -69% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 7% | 0% |
Purpose: Precise search, full-depth reading — precise investigation of a specific sub-problem. Not broad, not exhaustive — targeted and thorough.
When to use: User knows exactly what they're looking for and needs precise, detailed understanding. E.g., "How exactly does DPO handle the reward model?" or "What are all the variants of LoRA rank adaptation?"
| Base SOP | Target | ±10% Range | |----------|--------|------------| | web-search | 30 results | 27–33 | | web-research | 5 pages | 4–6 | | paper-overview | 40 papers | 36–44 | | 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 | 30 | ??? | ???% |
| web-research | 5 | ??? | ???% |
| paper-overview | 40 | ??? | ???% |
| paper-search | 40 | ??? | ???% |
| paper-research | 20 | ??? | ???% |Do not exit the strategy until all rows reach ≥90%.
None mandatory — CC composes directly from SOPs.
Import (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):
extract-data — structured comparison tables from deep-read papersgap-identification — find what the literature hasn't addressedsurvey-synthesis — produce final structured outputDetailed Technical Analysis containing:
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | extract-data | Structured data extraction from deep-read papers — produces comparison tables (method, dataset, metrics, results, limitations). Used by systematic-survey and deep-survey. | | 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. |
<!-- 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-05 | fail→pass | 14,568 | 2,364 | -84% | 1 | 1 | 0% | 736 | 1,503 | +104% | 0 | 0 | — |
case-01 | fail→pass | 30,972 | 5,029 | -84% | 1 | 1 | 0% | 6,216 | 1,846 | -70% | 0 | 0 | — |
case-02 | fail→pass | 33,351 | 5,969 | -82% | 1 | 1 | 0% | 6,208 | 1,946 | -69% | 0 | 0 | — |
case-03 | fail→fail | 34,986 | 34,747 | -1% | 1 | 1 | 0% | 6,204 | 7,304 | +18% | 0 | 0 | — |
case-04 | fail→pass | 12,955 | 2,563 | -80% | 1 | 1 | 0% | 1,978 | 1,495 | -24% | 0 | 0 | — |
case-06 | fail→pass | 10,593 | 3,457 | -67% | 1 | 1 | 0% | 1,588 | 1,707 | +7% | 0 | 0 | — |
case-07 | fail→pass | 8,287 | 2,729 | -67% | 1 | 1 | 0% | 1,148 | 1,532 | +33% | 0 | 0 | — |
case-08 | fail→pass | 8,180 | 2,096 | -74% | 1 | 1 | 0% | 1,282 | 1,433 | +12% | 0 | 0 | — |
case-09 | pass→pass | 4,439 | 2,367 | -47% | 1 | 1 | 0% | 592 | 1,458 | +146% | 0 | 0 | — |
case-10 | pass→pass | 8,197 | 4,860 | -41% | 1 | 1 | 0% | 1,174 | 1,834 | +56% | 0 | 0 | — |
case-11 | fail→pass | 5,487 | 2,290 | -58% | 1 | 1 | 0% | 739 | 1,487 | +101% | 0 | 0 | — |
case-12 | fail→pass | 15,634 | 1,936 | -88% | 1 | 1 | 0% | 1,080 | 1,408 | +30% | 0 | 0 | — |
case-13 | fail→fail | 11,378 | 2,472 | -78% | 1 | 1 | 0% | 1,847 | 1,453 | -21% | 0 | 0 | — |
case-14 | pass→pass | 9,586 | 4,611 | -52% | 1 | 1 | 0% | 1,333 | 1,837 | +38% | 0 | 0 | — |
case-15 | pass→pass | 8,026 | 3,868 | -52% | 1 | 1 | 0% | 1,135 | 1,633 | +44% | 0 | 0 | — |
case-16 | pass→pass | 7,817 | 2,956 | -62% | 1 | 1 | 0% | 1,167 | 1,582 | +36% | 0 | 0 | — |
case-17 | fail→pass | 16,426 | 9,483 | -42% | 1 | 1 | 0% | 2,399 | 2,609 | +9% | 0 | 0 | — |
case-18 | pass→pass | 17,495 | 13,420 | -23% | 1 | 1 | 0% | 2,737 | 3,227 | +18% | 0 | 0 | — |
case-19 | pass→pass | 8,735 | 1,886 | -78% | 1 | 1 | 0% | 1,345 | 1,337 | -1% | 0 | 0 | — |
case-20 | pass→pass | 16,264 | 12,232 | -25% | 1 | 1 | 0% | 2,609 | 3,045 | +17% | 0 | 0 | — |
case-21 | fail→fail | 8,050 | 5,678 | -29% | 1 | 1 | 0% | 708 | 1,944 | +175% | 0 | 0 | — |
case-22 | fail→fail | 6,351 | 10,757 | +69% | 1 | 1 | 0% | 1,019 | 2,675 | +163% | 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 20 counted toward the lift figure. The other 2 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 +45 percentage points is the difference between those two pass rates over the 20 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.