Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →13 deep research & systematic reviews skills. Trigger: systematic reviews, multi-source synthesis, comprehensive literature surveys. Design: multi-step research protocols with quality assessment and evidence grading.
.claude/skills/brycewang-stanford-deep-research-skills/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -31% | 0% |
Select the skill matching the user's need, then read its SKILL.md.
| Skill | Description | |-------|-------------| | auto-deep-research-guide | Automated deep research tool for thorough topic investigation | | deep-searcher-guide | Open deep research alternative for private data with vector search | | gpt-researcher-guide | Autonomous agent for comprehensive deep research on any topic | | in-depth-research-guide | Structured methodology for conducting exhaustive multi-source investigations | | khoj-research-guide | AI second brain for deep research and personal knowledge management | | kosmos-scientist-guide | Claude Code-driven autonomous AI Scientist for discovery | | llm-scientific-discovery-guide | Survey of LLM agents for biomedical scientific discovery | | local-deep-research-guide | Deep research agent searching 10+ sources with local or cloud LLMs | | meta-synthesis-guide | Conduct qualitative meta-synthesis and evidence synthesis methods | | open-researcher-guide | Open pipeline for generating deep research trajectories with LLMs | | scoping-review-guide | Scoping review methodology for broad evidence mapping | | systematic-review-guide | Systematic review methodology with PRISMA and evidence synthesis | | tongyi-deep-research-guide | Open-source deep research agent by Alibaba for scholarly research |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 46,332 | 52,603 | +14% | 1 | 1 | 0% | 3,181 | 1,147 | -64% | 0 | 0 | — |
case-02 | pass→pass | 52,166 | 20,787 | -60% | 1 | 1 | 0% | 3,069 | 4,065 | +32% | 0 | 0 | — |
case-03 | pass→fail | 14,160 | 5,379 | -62% | 1 | 1 | 0% | 2,548 | 738 | -71% | 0 | 0 | — |
case-04 | pass→fail | 42,502 | 8,901 | -79% | 1 | 1 | 0% | 1,920 | 1,157 | -40% | 0 | 0 | — |
case-05 | fail→pass | 16,362 | 5,473 | -67% | 1 | 1 | 0% | 2,400 | 990 | -59% | 0 | 0 | — |
case-06 | fail→pass | 13,719 | 3,788 | -72% | 1 | 1 | 0% | 2,235 | 1,258 | -44% | 0 | 0 | — |
case-07 | fail→fail | 17,345 | 5,692 | -67% | 1 | 1 | 0% | 2,803 | 985 | -65% | 0 | 0 | — |
case-08 | pass→pass | 13,275 | 5,920 | -55% | 1 | 1 | 0% | 2,206 | 1,482 | -33% | 0 | 0 | — |
case-09 | fail→pass | 19,465 | 6,550 | -66% | 1 | 1 | 0% | 3,468 | 1,798 | -48% | 0 | 0 | — |
case-10 | fail→pass | 15,934 | 2,849 | -82% | 1 | 1 | 0% | 2,546 | 1,104 | -57% | 0 | 0 | — |
case-11 | fail→fail | 9,639 | 6,213 | -36% | 1 | 1 | 0% | 1,388 | 913 | -34% | 0 | 0 | — |
case-12 | fail→pass | 11,576 | 3,974 | -66% | 1 | 1 | 0% | 1,844 | 1,280 | -31% | 0 | 0 | — |
case-13 | fail→fail | 19,750 | 6,602 | -67% | 1 | 1 | 0% | 3,119 | 1,016 | -67% | 0 | 0 | — |
case-14 | pass→fail | 19,267 | 4,304 | -78% | 1 | 1 | 0% | 3,450 | 901 | -74% | 0 | 0 | — |
case-15 | pass→pass | 9,236 | 8,598 | -7% | 1 | 1 | 0% | 1,802 | 2,124 | +18% | 0 | 0 | — |
case-16 | pass→pass | 8,842 | 5,908 | -33% | 1 | 1 | 0% | 1,847 | 1,686 | -9% | 0 | 0 | — |
case-17 | pass→fail | 10,140 | 5,732 | -43% | 1 | 1 | 0% | 1,807 | 846 | -53% | 0 | 0 | — |
case-18 | pass→pass | 22,375 | 10,855 | -51% | 1 | 1 | 0% | 3,454 | 2,291 | -34% | 0 | 0 | — |
case-19 | pass→pass | 9,686 | 9,024 | -7% | 1 | 1 | 0% | 1,462 | 1,339 | -8% | 0 | 0 | — |
case-20 | fail→pass | 15,160 | 5,706 | -62% | 1 | 1 | 0% | 2,267 | 1,272 | -44% | 0 | 0 | — |
case-21 | fail→fail | 12,651 | 6,922 | -45% | 1 | 1 | 0% | 2,216 | 956 | -57% | 0 | 0 | — |
case-22 | fail→pass | 14,285 | 4,414 | -69% | 1 | 1 | 0% | 2,332 | 1,226 | -47% | 0 | 0 | — |
case-23 | fail→fail | 18,290 | 7,295 | -60% | 1 | 1 | 0% | 2,579 | 915 | -65% | 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. 23 cases were attempted, and 13 counted toward the lift figure. The other 10 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 +13 percentage points is the difference between those two pass rates over the 13 comparable cases. 4 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.