Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Deep Insight Engine with 5 campaigns (gap-analysis, insight, boundary-analysis, sensitivity-analysis, problem-reformulation). Use this skill whenever a user needs to deeply analyze research gaps, understand root causes, probe method boundaries, assess assumption sensitivity, or reformulate research problems. Pre-condition: north-star-crystallization complete + knowledge-acquisition campaign has produced initial findings.
.claude/skills/yogsoth-ai-deep-insight/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 101% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 141% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 1861% | 0% |
Deep insight engine — from surface phenomena to root causes, boundaries, assumptions, and the problem itself. Five campaigns, each a self-contained autonomous analysis domain. You provide a research gap or finding — the engine routes to the right campaign, selects a strategy, and executes autonomously with quantitative budget enforcement.
Strategy Book mode. This file is a textbook, not a script. CC reads, internalizes principles, then autonomously constructs the analysis approach for the specific research situation.
Hard constraints only:
Everything else — execution order, iteration count, tactic selection, SOP combination — is CC's autonomous decision.
ENTRY.md (this file)
→ Campaign (5): self-contained deep analysis domain
→ Strategy: selected by analysis purpose/intent
→ Tactic: multi-step orchestration pattern (reusable across strategies)
→ SOP: single operation (import or subagent)CC can skip the tactic layer and use SOPs directly when the task is simple enough.
| Signal | Campaign | |--------|----------| | gap identification, gap classification, evidence mapping, prioritization, gap validation | → gap-analysis | | root-cause analysis, stakeholders, tensions, HMW, assumption audit, 5 Whys | → insight | | validity boundaries, method failure, robustness, distribution shift, scaling limits | → boundary-analysis | | assumption ranking, sensitivity, variance decomposition, uncertainty propagation, critical path | → sensitivity-analysis | | redefining the problem, dominant ideas, multiple perspectives, double-loop learning, wicked problems | → problem-reformulation |
Campaigns can be composed:
The orchestrator decides composition based on the research state and user intent.
| MCP Server | Tools | |------------|-------| | brave-search | brave_web_search, brave_news_search, brave_llm_context | | apify | rag-web-browser, google-scholar-scraper | | alphaxiv | discover_papers, get_paper_content, answer_pdf_queries, read_files_from_github_repository | | semantic-scholar | ss_paper, ss_paper_batch, ss_references, ss_citations, ss_recommendations, ss_relevance_search, ss_author, ss_author_papers |
context/deep-insight-[campaign]-[topic].md| Dependency | What It Provides | |-----------|-----------------| | web-browsing | web-search + web-research | | literature-engine | literature-overview + literature-search + literature-research | | subagent-spawning | Subagent dispatch conventions | | context-management | Checkpoint protocol | | north-star-crystallization | Pre-condition (research intent) | | knowledge-acquisition | Pre-condition (initial findings) |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Campaign | When to use | | --- | --- | | boundary-analysis | Boundary Analysis Campaign — probe where methods fail, map validity envelopes, test robustness, catalog failure modes, detect scaling limits. 5 strategies, 3 tactics, 11 subagent SOPs. | | deep-insight-sensitivity-analysis | Sensitivity Analysis Campaign — identify which assumptions are most critical by measuring their impact on conclusions. 5 strategies (parameter-screening, variance-decomposition, assumption-criticality, uncertainty-propagation, decision-sensitivity), 3 tactics, 11 subagent SOPs. | | gap-analysis | Gap Analysis Campaign — identify, classify, validate, and prioritize research gaps via systematic evidence mapping. 5 strategies (gap-identification, gap-classification, gap-validation, gap-prioritization, gap-synthesis), 3 tactics, 12 subagent SOPs. | | insight | Insight Campaign — deep root-cause analysis of why research gaps persist. 5 strategies (root-cause-drilling, stakeholder-mapping, tension-mining, question-reformulation, assumption-audit), 4 tactics, 13 subagent SOPs. | | problem-reformulation | Problem Reformulation Campaign — question the problem itself. Escape dominant ideas, reframe from multiple perspectives, apply dialectical inquiry, assess wickedness, discover appreciative alternatives. 5 strategies, 3 tactics, 10 subagent SOPs. |
<!-- 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-11 | pass→pass | 12,902 | 7,838 | -39% | 1 | 1 | 0% | 1,900 | 2,438 | +28% | 0 | 0 | — |
case-12 | pass→pass | 19,462 | 13,572 | -30% | 1 | 1 | 0% | 2,930 | 3,361 | +15% | 0 | 0 | — |
case-04 | fail→fail | 16,644 | 8,945 | -46% | 1 | 1 | 0% | 2,418 | 2,110 | -13% | 0 | 0 | — |
case-05 | pass→pass | 20,242 | 30,317 | +50% | 1 | 1 | 0% | 3,327 | 6,425 | +93% | 0 | 0 | — |
case-01 | fail→fail | 34,930 | 13,702 | -61% | 1 | 1 | 0% | 5,106 | 2,325 | -54% | 0 | 0 | — |
case-02 | fail→fail | 94,288 | 38,052 | -60% | 1 | 1 | 0% | 5,176 | 7,439 | +44% | 0 | 0 | — |
case-03 | fail→fail | 25,610 | 39,619 | +55% | 1 | 1 | 0% | 3,779 | 7,440 | +97% | 0 | 0 | — |
case-06 | fail→pass | 19,765 | 33,713 | +71% | 1 | 1 | 0% | 3,118 | 6,255 | +101% | 0 | 0 | — |
case-07 | pass→pass | 22,540 | 23,039 | +2% | 1 | 1 | 0% | 3,434 | 4,453 | +30% | 0 | 0 | — |
case-08 | pass→pass | 31,101 | 35,194 | +13% | 1 | 1 | 0% | 5,643 | 7,325 | +30% | 0 | 0 | — |
case-09 | fail→fail | 38,171 | 81,560 | +114% | 1 | 1 | 0% | 6,195 | 7,425 | +20% | 0 | 0 | — |
case-10 | fail→pass | 15,530 | 11,905 | -23% | 1 | 1 | 0% | 2,388 | 3,017 | +26% | 0 | 0 | — |
case-13 | fail→pass | 10,036 | 5,171 | -48% | 1 | 1 | 0% | 1,466 | 2,045 | +39% | 0 | 0 | — |
case-14 | fail→fail | 10,757 | 4,509 | -58% | 1 | 1 | 0% | 1,900 | 1,995 | +5% | 0 | 0 | — |
case-15 | fail→pass | 20,096 | 40,123 | +100% | 1 | 1 | 0% | 3,013 | 7,254 | +141% | 0 | 0 | — |
case-16 | pass→pass | 21,886 | 44,306 | +102% | 1 | 1 | 0% | 3,456 | 7,294 | +111% | 0 | 0 | — |
case-17 | pass→pass | 21,574 | 28,606 | +33% | 1 | 1 | 0% | 3,775 | 5,834 | +55% | 0 | 0 | — |
case-18 | pass→fail | 30,350 | 6,268 | -79% | 1 | 1 | 0% | 6,010 | 1,588 | -74% | 0 | 0 | — |
case-19 | pass→pass | 11,717 | 14,235 | +21% | 1 | 1 | 0% | 2,277 | 4,173 | +83% | 0 | 0 | — |
case-20 | fail→pass | 2,817 | 29,895 | +961% | 1 | 1 | 0% | 369 | 7,235 | +1861% | 0 | 0 | — |
case-21 | fail→pass | 25,632 | 2,346 | -91% | 1 | 1 | 0% | 1,521 | 1,679 | +10% | 0 | 0 | — |
case-22 | fail→pass | 13,808 | 7,499 | -46% | 1 | 1 | 0% | 2,129 | 2,512 | +18% | 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 19 counted toward the lift figure. The other 3 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 +27 percentage points is the difference between those two pass rates over the 19 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.