Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Synthesize research findings from memory into structured reports with evidence grading, contradiction resolution, and actionable recommendations
.claude/skills/ruvnet-research-synthesize/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -35% | 0% |
Synthesize accumulated research findings into actionable reports.
After running deep-research (one or multiple times), when you need to pull together findings from memory into a coherent synthesis with recommendations.
mcp__plugin_ruflo-core_ruflo__memory_search namespace research for raw findingsmcp__plugin_ruflo-core_ruflo__memory_search namespace research-sources for referencesmcp__plugin_ruflo-core_ruflo__agentdb_pattern-search for discovered patternsmcp__plugin_ruflo-core_ruflo__agentdb_context-synthesize for AI-assisted context buildingmcp__plugin_ruflo-core_ruflo__neural_predict to score which findings are most relevant to the original goalmcp__plugin_ruflo-core_ruflo__memory_store namespace research-synthesis with the full report# [Research Topic] — Synthesis Report
## Summary
[2-3 sentence answer]
## Key Findings
1. [Finding] — Evidence: High/Medium/Low
2. [Finding] — Evidence: High/Medium/Low
## Contradictions
- [Claim A] vs [Claim B]: [resolution or "unresolved"]
## Recommendations
1. [Action] — because [reasoning]
## Sources
- [key]: [description]| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-10 | fail→pass | 11,424 | 2,683 | -77% | 1 | 1 | 0% | 2,103 | 1,148 | -45% | 0 | 0 | — |
case-01 | fail→fail | 27,201 | 6,755 | -75% | 1 | 1 | 0% | 4,770 | 1,058 | -78% | 0 | 0 | — |
case-02 | fail→fail | 26,295 | 5,574 | -79% | 1 | 1 | 0% | 4,857 | 997 | -79% | 0 | 0 | — |
case-03 | fail→fail | 37,089 | 6,820 | -82% | 1 | 1 | 0% | 6,198 | 928 | -85% | 0 | 0 | — |
case-04 | fail→pass | 11,456 | 3,539 | -69% | 1 | 1 | 0% | 1,887 | 1,029 | -45% | 0 | 0 | — |
case-05 | pass→pass | 4,112 | 2,611 | -37% | 1 | 1 | 0% | 651 | 1,013 | +56% | 0 | 0 | — |
case-06 | fail→pass | 9,294 | 3,102 | -67% | 1 | 1 | 0% | 1,545 | 1,015 | -34% | 0 | 0 | — |
case-07 | fail→pass | 10,418 | 5,759 | -45% | 1 | 1 | 0% | 1,784 | 1,506 | -16% | 0 | 0 | — |
case-08 | pass→pass | 11,380 | 4,168 | -63% | 1 | 1 | 0% | 1,796 | 1,218 | -32% | 0 | 0 | — |
case-09 | fail→fail | 9,384 | 2,847 | -70% | 1 | 1 | 0% | 1,705 | 1,086 | -36% | 0 | 0 | — |
case-11 | fail→fail | 3,509 | 1,312 | -63% | 1 | 1 | 0% | 554 | 775 | +40% | 0 | 0 | — |
case-12 | fail→pass | 13,716 | 6,597 | -52% | 1 | 1 | 0% | 2,459 | 1,607 | -35% | 0 | 0 | — |
case-13 | fail→fail | 7,624 | 2,975 | -61% | 1 | 1 | 0% | 1,270 | 1,057 | -17% | 0 | 0 | — |
case-14 | fail→fail | 8,048 | 3,242 | -60% | 1 | 1 | 0% | 1,292 | 1,119 | -13% | 0 | 0 | — |
case-15 | fail→pass | 11,013 | 8,470 | -23% | 1 | 1 | 0% | 1,885 | 1,508 | -20% | 0 | 0 | — |
case-16 | pass→pass | 10,353 | 4,190 | -60% | 1 | 1 | 0% | 1,745 | 1,299 | -26% | 0 | 0 | — |
case-17 | pass→fail | 9,422 | 2,656 | -72% | 1 | 1 | 0% | 1,638 | 1,013 | -38% | 0 | 0 | — |
case-18 | pass→fail | 10,615 | 3,256 | -69% | 1 | 1 | 0% | 1,680 | 1,092 | -35% | 0 | 0 | — |
case-19 | pass→pass | 5,120 | 9,534 | +86% | 1 | 1 | 0% | 1,029 | 2,110 | +105% | 0 | 0 | — |
case-20 | pass→fail | 16,682 | 4,849 | -71% | 1 | 1 | 0% | 3,073 | 905 | -71% | 0 | 0 | — |
case-21 | pass→pass | 13,434 | 8,634 | -36% | 1 | 1 | 0% | 2,783 | 2,509 | -10% | 0 | 0 | — |
case-22 | pass→pass | 10,593 | 4,098 | -61% | 1 | 1 | 0% | 1,769 | 1,208 | -32% | 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 18 counted toward the lift figure. The other 4 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 18 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.