// Complete strategy book for the research engine. Contains all campaign descriptions, routing tables, pre-conditions, and orchestration rules. CC reads this before generating any Research Spec.
Structured questioning SOP to determine which campaigns to include, emphasize, or skip. Used during spec generation.
Structured questioning SOP to identify practical constraints that shape the research spec. Used during spec generation.
Structured questioning SOP to determine research boundaries, depth, and breadth. Used during spec generation.
Quality gate for Research Specs. Checks for placeholders, consistency, scope, ambiguity, context protocol, and quantification. Mandatory before user review.
Top-level orchestrator for the yogsoth-ai research ecosystem. Drives the full research lifecycle from direction crystallization through experiment design.
Execute a Research Spec step by step, respecting context protocol, deviation limits, and backtrack rules. Supports multi-session recovery.
| name | research-catalog |
| description | Complete strategy book for the research engine. Contains all campaign descriptions, routing tables, pre-conditions, and orchestration rules. CC reads this before generating any Research Spec. |
| execution | reference |
This is the strategy book for the De-Anthropocentric Research Engine. Read this document in full before generating any Research Spec.
north-star-crystallization
→ knowledge-acquisition
→ deep-insight
→ hypothesis-formation
→ creative-ideation
→ convergence
→ stress-test
→ experiment-execution (design only)
Each campaign can feed back to earlier stages (see backtrack conditions in each section).
Transform fuzzy research intent into a crystallized North Star and structured ResearchBrief.
You are a Goal-Driven Requirement Refinement Engine. Through adaptive dialogue and on-demand investigation (web search, paper search), you help users who range from "I have no idea what to research" to "I have a specific topic but need structure" arrive at:
Read the user's first message and assess their information density:
| Signal | Route to |
|---|---|
| No direction at all ("I want to publish but don't know what") | cold-start |
| Has a general direction but not specific ("I'm interested in LLM reasoning") | warm-start |
| Has a specific topic/problem ("I want to improve CoT faithfulness") | hot-start |
ENTRY.md (this file)
鈫?Campaign (1): north-star-crystallization
鈫?Strategy (3): cold-start, warm-start, hot-start
鈫?Tactic (6): actor-profiling, landscape-reconnaissance, direction-narrowing,
obstacle-analysis, goal-decomposition, north-star-synthesis
鈫?SOP (23): dialogue + investigation operations
This is a strategy book, not a pipeline orchestration file. Strategies provide war doctrine and available tactics. Tactics provide methodology guidance and available SOPs. SOPs are specific techniques. You are the general 鈥?read the book, then decide.
This skill is a pre-condition for knowledge-acquisition. All 5 knowledge-acquisition campaigns require a crystallized North Star before execution.
| MCP Server | Tools |
|---|---|
| brave-search | brave_web_search, brave_llm_context |
| apify | rag-web-browser, google-scholar-scraper |
| alphaxiv | discover_papers, get_paper_content |
| semantic-scholar | ss_relevance_search, ss_paper |
| Dependency | What It Provides |
|---|---|
| web-browsing | web-search + web-research |
| literature-engine | paper-overview + paper-search |
| subagent-spawning | Subagent dispatch conventions |
Systematic research knowledge acquisition engine. Five campaigns, each a self-contained autonomous research activity domain. You provide a research intent 鈥?the engine routes to the right campaign, selects a strategy, and executes autonomously with quantitative budget enforcement.
North-star-crystallization must be complete before entering any campaign. Research intent must be fully crystallized.
ENTRY.md (this file)
鈫?Campaign (5): self-contained research activity domain
鈫?Strategy: selected by analysis purpose/intent
鈫?Tactic: multi-step orchestration pattern (reusable across strategies)
鈫?SOP: single operation (import or subagent)
| Signal | Campaign |
|---|---|
| 鏂囩尞璋冪爺銆佺患杩般€佽鏂囨悳绱€丳RISMA銆乻nowball | 鈫?literature-survey |
| 涓撳埄鍒嗘瀽銆乸rior art銆佺櫧绌洪棿銆佹潈鍒╄姹傘€両PC | 鈫?patent-mining |
| benchmark 鍒嗘瀽銆佽瘎浼版柟娉曘€乵etric 缂洪櫡銆佹帓琛屾銆侀ケ鍜屽害 | 鈫?benchmark-archaeology |
| 璺ㄧ爺绌剁粺璁＄患鍚堛€佹晥搴旈噺銆佸紓璐ㄦ€с€佸彂琛ㄥ亸鍊氥€丟RADE | 鈫?meta-analysis |
| SOTA 鏁寸悊銆佹€ц兘瀵规瘮銆乥aseline 澶嶇幇銆佽繘灞曟洸绾? | 鈫?baseline-establishment |
Campaigns can be composed:
The orchestrator decides composition based on the crystallized North Star statement.
| 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 |
| 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 |
Deep insight engine 鈥?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 鈥?the engine routes to the right campaign, selects a strategy, and executes autonomously with quantitative budget enforcement.
鍏垫硶涔?(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 鈥?execution order, iteration count, tactic selection, SOP combination 鈥?is CC's autonomous decision.
ENTRY.md (this file)
鈫?Campaign (5): self-contained deep analysis domain
鈫?Strategy: selected by analysis purpose/intent
鈫?Tactic: multi-step orchestration pattern (reusable across strategies)
鈫?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 璇嗗埆銆佺┖鐧藉垎绫汇€佽瘉鎹湴鍥俱€佷紭鍏堢骇鎺掑簭銆乬ap 楠岃瘉 | 鈫?gap-analysis |
| 鏍瑰洜鍒嗘瀽銆佸埄鐩婄浉鍏宠€呫€佸紶鍔涖€丠MW銆佸亣璁惧璁°€? Whys | 鈫?insight |
| 鏈夋晥鎬ц竟鐣屻€佹柟娉曞け鏁堛€侀瞾妫掓€с€佸垎甯冨亸绉汇€佽妯℃瀬闄? | 鈫?boundary-analysis |
| 鍋囪鎺掑簭銆佹晱鎰熸€с€佹柟宸垎瑙ｃ€佷笉纭畾鎬т紶鎾€佸叧閿矾寰? | 鈫?sensitivity-analysis |
| 閲嶆柊瀹氫箟闂銆佷富瀵艰蹇点€佸瑙嗚銆佸弻鐜涔犮€侀偑鎭堕棶棰? | 鈫?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) |
Goal-Driven Hypothesis & Research Question Formation Engine 鈥?灏嗕笂娓哥殑 gaps 鍜?insights 杞寲涓哄彲娴嬭瘯鐨勫亣璁惧拰绮剧‘鐨勭爺绌堕棶棰樸€?
鍓嶇疆鏉′欢:
| Campaign | 鏍稿績闂 | 杈撳叆 | 浜у嚭 |
|---|---|---|---|
| gap-prioritization | "鍝簺 gap 鏈€鍊煎緱鏀诲嚮锛? | 涓婃父 gaps | 鎺掑簭鍚庣殑 gap 浼樺厛绾у垪琛?+ 鏀诲嚮寤鸿 |
| hypothesis-formulation | "濡備綍灏?insight 杞寲涓哄彲娴嬭瘯鍋囪锛? | gaps + insights + tensions | 缁撴瀯鍖栧亣璁?+ falsifiability criteria |
| research-question | "濡備綍灏嗗亣璁剧粏鍖栦负绮剧‘鐮旂┒闂锛? | 鍋囪 + 棰嗗煙绾︽潫 | 妗嗘灦鍖栫殑鐮旂┒闂 + scope + success criteria |
| Signal | Campaign |
|---|---|
| gap 鎺掑簭銆佷紭鍏堢骇銆佸摢涓€煎緱鍋氥€丳iCMe銆佸缁磋瘎鍒嗐€乸ortfolio | 鈫?gap-prioritization |
| 鍋囪鐢熸垚銆佺悊璁烘帹瀵笺€佸彲璇佷吉銆両f-then銆佸彉閲忋€佹満鍒躲€乧ompeting hypothesis | 鈫?hypothesis-formulation |
| 鐮旂┒闂銆丳ICO銆丼PIDER銆丗INER銆乻cope銆佸瓙闂鍒嗚В銆乻uccess criteria | 鈫?research-question |
涓変釜 campaign 瀹屽叏鐏垫椿缁勫悎锛孋C 鑷富鍐冲畾鐢ㄥ嚑涓€佷粈涔堥『搴忥細
鍏垫硶涔︽ā寮?鈥?CC 璇诲畬鍚庡唴鍖栧師鍒欙紝闈㈠鍏蜂綋鐮旂┒浠诲姟鑷鏋勫缓鎵撴硶銆? 纭害鏉熶粎鍥涚:
name: creative-ideation description: Creative Generation Engine 鈥?transforms research hypotheses into diverse solution spaces via 10 parallel creativity campaigns spanning structural, analogical, destructive, and combinatorial methods. execution: entry pre-conditions:
| Signal Keywords | Route To |
|---|---|
| SCAMPER, TRIZ, 缁勪欢鎵嬫湳, 缁撴瀯鍙樻崲, 鍔熻兘瑁佸壀 | 鈫?structural-deconstruction |
| 璺ㄥ煙, 绫绘瘮杩佺Щ, bisociation, 闅忔満鍒烘縺, 寮哄埗杩炴帴 | 鈫?cross-domain-discovery |
| 鍋囪鍚﹀畾, 鍙嶅悜澶磋剳椋庢毚, 鏈€宸柟娉? 鍙?benchmark | 鈫?assumption-destruction |
| 浠跨敓, 鐢熺墿绫绘瘮, 鑷劧绛栫暐, BioTRIZ, 鐢熸€佹ā寮? | 鈫?biomimicry |
| 绫绘瘮, 闅愬柣, 杩滆冻娉? 涓汉绫绘瘮, 绗﹀彿鍘嬬缉 | 鈫?synectics |
| 褰㈡€佸垎鏋? Zwicky box, CCA, 缁村害缁勫悎, 璁捐绌洪棿 | 鈫?morphological-exploration |
| PO, 妯悜鎬濈淮, 姒傚康鎵? 闅忔満鍏ュ彛, 鎸戞垬鎿嶄綔 | 鈫?lateral-thinking |
| 姒傚康铻嶅悎, blending, 娑岀幇, 澶氬眰缁勫悎, 鍔熻兘閲嶅垎閰? | 鈫?combinatorial-creativity |
| 瑙嗚鍒囨崲, 鍏附, 瑙掕壊鎵紨, 绾︽潫娉ㄥ叆, 鏃堕棿鎶曞皠 | 鈫?perspective-forcing |
| 绌蜂妇, 瑕嗙洊鍒嗘瀽, 鏂规硶鐭╅樀, ablation, 澶辫触鍒嗙被 | 鈫?systematic-enumeration |
When the research problem warrants broad creative exploration, CC may invoke multiple campaigns in parallel. Four natural cluster families:
| Cluster | Campaigns | When |
|---|---|---|
| 绫绘瘮鏃?(Analogy) | cross-domain-discovery + biomimicry + synectics | Problem benefits from external domain transfer |
| 缁勫悎鏃?(Combinatorial) | structural-deconstruction + morphological-exploration + combinatorial-creativity | Problem has decomposable structure |
| 棰犺鏃?(Disruptive) | assumption-destruction + lateral-thinking + perspective-forcing | Problem is stuck in dominant paradigm |
| 瑕嗙洊鏃?(Coverage) | systematic-enumeration + morphological-exploration | Need exhaustive space mapping |
鍏ㄩ潰鍙戞暎: Invoke 3-5 campaigns based on problem characteristics. Each campaign executes independently with its own context file. Results aggregated at ENTRY level.
CC decides:
ENTRY.md (this file)
鈹斺攢鈹€ Campaign (10) 鈥?self-contained creative activity domain
鈹斺攢鈹€ Strategy (5 per campaign) 鈥?iterative framework with budget + state ledger
鈹斺攢鈹€ Tactic (2-3 per campaign) 鈥?SOP combination principle
鈹斺攢鈹€ SOP 鈥?single operation (subagent or import)
| Server | Tools |
|---|---|
| brave-search | brave_web_search, brave_news_search, brave_llm_context |
| apify | rag-web-browser |
| alphaxiv | discover_papers, get_paper_content, answer_pdf_queries |
| semantic-scholar | paper, paperBatch, references, citations, recommendations, relevanceSearch |
| Dependency | What It Provides |
|---|---|
| web-browsing | web-search + web-research SOPs |
| literature-engine | paper-overview + paper-search + paper-research SOPs |
| subagent-spawning | Subagent dispatch conventions (spawn-agent skill) |
| context-management | Checkpoint protocol (context-init, context-checkpoint) |
context-init 鈥?load or create campaign context filecontext-checkpoint 鈥?append strategy output to campaign fileContext file naming: context/creative-ideation-[campaign]-[topic].md
This engine STOPS at idea generation. Its output is a structured set of diverse ideas with:
It does NOT:
| SOP | Source | Quality Gate |
|---|---|---|
| web-search | web-browsing | Snippets only 鈥?no conclusions from snippets |
| web-research | web-browsing | Must fetch full page via apify |
| paper-overview | literature-engine | Abstract only 鈥?no methodology conclusions |
| paper-search | literature-engine | AI summary 鈥?sufficient for methodology understanding |
| paper-research | literature-engine | Full text 鈥?required for quoting results |
| SOP | Used By |
|---|---|
| saturation-detection | All 10 campaigns |
| novelty-scoring | All 10 campaigns |
| idea-synthesis | All 10 campaigns |
| domain-scanning | cross-domain, biomimicry, synectics, combinatorial |
| assumption-surfacing | assumption-destruction, lateral-thinking, perspective-forcing, structural-deconstruction |
| constraint-injection | perspective-forcing, lateral-thinking, structural-deconstruction, morphological-exploration |
| parameter-identification | structural-deconstruction, morphological-exploration, combinatorial-creativity, systematic-enumeration |
| po-provocation | assumption-destruction, lateral-thinking, perspective-forcing |
| random-word-stimulus | cross-domain-discovery, lateral-thinking, synectics |
| Tactic | Used By |
|---|---|
| analogy-extraction | cross-domain-discovery, synectics, biomimicry |
| combination-mapping | morphological-exploration, combinatorial-creativity, structural-deconstruction, systematic-enumeration |
| provocation-generation | lateral-thinking, assumption-destruction, perspective-forcing |
| evaluation-filtering | All 10 campaigns |
Universal convergence engine. Six campaigns, each a self-contained convergence paradigm. You provide a candidate set and a convergence intent 鈥?the engine routes to the right campaign, selects a strategy, and executes autonomously with quantitative budget enforcement.
ENTRY.md (this file)
鈫?Campaign (6): self-contained convergence paradigm
鈫?Strategy: selected by convergence intent/scenario
鈫?Tactic: multi-step orchestration pattern
鈫?SOP: single operation (import or subagent)
| Signal | Campaign |
|---|---|
| score/rank candidates against multiple criteria | 鈫?multi-criteria-scoring |
| produce global ranking via pairwise comparisons | 鈫?pairwise-ranking |
| multiple perspectives disagree, need convergence | 鈫?structured-consensus |
| assess feasibility/readiness of candidates | 鈫?feasibility-assessment |
| select a balanced portfolio from candidates | 鈫?portfolio-optimization |
| verify rejected candidates, stress-test winners | 鈫?steel-manning |
| criteria themselves are contested | 鈫?structured-consensus 鈫?multi-criteria-scoring |
| high-stakes decision requiring method robustness | 鈫?multi-criteria-scoring (multi-method-triangulation) |
| too many candidates, need coarse screening first | 鈫?multi-criteria-scoring (non-compensatory-screening) 鈫?pairwise-ranking |
CC has full autonomy to compose campaigns:
| Campaign | Output Type |
|---|---|
| multi-criteria-scoring | RankedList[] 鈥?scored and ranked candidates with scores, ranks, confidence |
| pairwise-ranking | RankedList[] 鈥?global ranking with ratings, convergence, consistency |
| structured-consensus | ConsensusReport 鈥?agreement report with consensus level, disagreement points, argument structure |
| feasibility-assessment | FeasibilityMatrix 鈥?readiness matrix with dimension scores, blockers, paths |
| portfolio-optimization | Portfolio 鈥?selected combination with coverage, risk distribution, budget allocation |
| steel-manning | SteelManVerdict 鈥?adversarial verification verdict (ACCEPT / REJECT / REVISE) |
| MCP Server | Tools |
|---|---|
| brave-search | brave_web_search, brave_llm_context |
| apify | rag-web-browser |
| alphaxiv | discover_papers, get_paper_content, answer_pdf_queries |
| semantic-scholar | relevanceSearch, paper, citations |
| wiki-vault | vault_search, vault_add_edge, vault_query_graph |
| Dependency | What It Provides |
|---|---|
| web-browsing | web-search + web-research (import SOPs) |
| literature-engine | paper-overview + paper-search + paper-research (import SOPs) |
| subagent-spawning | Subagent dispatch conventions (spawn-agent) |
| context-management | context-init + context-checkpoint |
| wiki-vault MCP | Knowledge persistence to graph |
Research Artifact Stress-Testing Engine 鈥?from structured debate to logical extreme, every claim must survive or be annotated. Five campaigns, each a self-contained adversarial validation domain. You provide a research artifact 鈥?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 validation approach for the specific artifact.
Hard constraints only:
Everything else 鈥?execution order, iteration count, tactic selection, SOP combination 鈥?is CC's autonomous decision.
| Type | Description |
|---|---|
gap | Research gap |
hypothesis | Testable hypothesis |
research-question | Research question |
idea | Creative solution |
approach | Selected method path |
experiment-design | Experiment design |
claim | Any research claim |
ENTRY.md (this file)
鈫?Campaign (5): self-contained adversarial validation domain
鈫?Strategy: selected by validation purpose/intent
鈫?Tactic: multi-step orchestration pattern (reusable across strategies)
鈫?SOP: single operation (import or subagent)
CC can skip the tactic layer and use SOPs directly when the task is simple enough.
| Signal | Campaign |
|---|---|
| Adversarial debate, multi-perspective review | 鈫?multiagent-debate |
| Systematic attack, assumption challenge | 鈫?red-teaming |
| Failure mode prediction, risk assessment | 鈫?failure-anticipation |
| Critical dependency probing, causal necessity | 鈫?counterfactual-probing |
| Logical falsification, boundary testing | 鈫?adversarial-stress-testing |
Campaigns can be composed:
The orchestrator decides composition based on artifact type and validation needs.
| MCP Server | Tools | Purpose |
|---|---|---|
| brave-search | brave_web_search, brave_news_search, brave_llm_context | Web discovery |
| apify | rag-web-browser | Full-text web retrieval |
| alphaxiv | discover_papers, get_paper_content, answer_pdf_queries | Paper access |
| semantic-scholar | relevanceSearch, paper, paperBatch, citations, references | Paper metadata |
context-init 鈥?initialize context filecontext-checkpoint 鈥?append 鈮?00 linesEach campaign produces its own typed report:
DebateVerdict (multiagent-debate)RedTeamReport (red-teaming)FailureAnticipationReport (failure-anticipation)CounterfactualMap (counterfactual-probing)AdversarialStressReport (adversarial-stress-testing)verdict-synthesis SOP can aggregate multiple campaign results into a unified StressTestSummary.
| Dependency | Provides |
|---|---|
| web-browsing | web-search, web-research (Import SOP) |
| literature-engine | paper-overview, paper-search, paper-research (Import SOP) |
| subagent-spawning | spawn-agent (execution runtime) |
| context-management | context-init, context-checkpoint (state persistence) |
| deep-insight | assumption-surfacing, evidence-synthesis, multi-stakeholder-simulation (cross-repo shared SOP) |
Four-Campaign Experiment Execution Engine 鈥?starting from validated research hypotheses/approaches, completes the full pipeline of experiment design, constraint analysis, scenario planning, implementation planning, and execution.
Prerequisites:
Execution boundary: Full pipeline 鈥?from experiment design to actual execution to result collection and analysis.
Outputs:
| Campaign | Core Question | Input | Output |
|---|---|---|---|
| experiment-design | "What experiment to run?" | Validated hypotheses/approaches | Complete experiment design |
| constraint-analysis | "What limits us?" | Experiment design + real-world constraints | Constraint analysis report + resolution plans |
| scenario-planning | "What might the future look like?" | Research approach + uncertainties | Multi-scenario analysis + robustness assessment |
| implementation-planning | "How to do it + do it" | Design + constraints + scenarios | Executable plan + execution results |
| Signal | Campaign |
|---|---|
| experiment design, factors, variables, ablation, baseline comparison, statistical methods | 鈫?experiment-design |
| bottleneck, constraint, insufficient resources, dependencies, conflicts | 鈫?constraint-analysis |
| scenarios, future, robustness, worst case, competitors, timeline | 鈫?scenario-planning |
| planning, execution, implementation, running experiments, result analysis, reproducibility | 鈫?implementation-planning |
The four campaigns can be flexibly combined; CC autonomously decides how many to use and in what order:
Art of War mode 鈥?CC internalizes the principles after reading, then autonomously constructs its approach for each specific research task.
Only four hard constraints:
CC has autonomous decision authority over:
web-search: Quick factual lookups, specific queriesweb-research: Deep multi-source investigationliterature-overview: High-level landscape surveyliterature-search: Targeted paper discoveryliterature-research: Deep paper analysis and synthesisspawn-agent: Dispatch parallel research subagentscontext-init: Create new context file for a campaigncontext-checkpoint: Append ≥500 lines of process + results to current context file