Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Creative Generation Engine — transforms research hypotheses into diverse solution spaces via 10 parallel creativity campaigns spanning structural, analogical, destructive, and combinatorial methods.
.claude/skills/yogsoth-ai-creative-ideation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 172% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 134% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 220% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -19% | 0% |
| Signal Keywords | Route To | |----------------|----------| | SCAMPER, TRIZ, component surgery, structural transformation, function trimming | → structural-deconstruction | | Cross-domain, analogical transfer, bisociation, random stimulus, forced connection | → cross-domain-discovery | | Assumption negation, reverse brainstorming, worst method, anti-benchmark | → assumption-destruction | | Biomimicry, biological analogy, natural strategies, BioTRIZ, ecosystem patterns | → biomimicry | | Analogy, metaphor, excursion method, personal analogy, symbolic compression | → synectics | | Morphological analysis, Zwicky box, CCA, dimension combination, design space | → morphological-exploration | | PO, lateral thinking, concept fan, random entry, challenge operation | → lateral-thinking | | Concept blending, blending, emergence, multi-level combination, function redistribution | → combinatorial-creativity | | Perspective switching, six hats, role play, constraint injection, temporal projection | → perspective-forcing | | Enumeration, coverage analysis, method matrix, ablation, failure taxonomy | → systematic-enumeration |
When the research problem warrants broad creative exploration, CC may invoke multiple campaigns in parallel. Four natural cluster families:
| Cluster | Campaigns | When | |---------|-----------|------| | Analogy family (Analogy) | cross-domain-discovery + biomimicry + synectics | Problem benefits from external domain transfer | | Combinatorial family (Combinatorial) | structural-deconstruction + morphological-exploration + combinatorial-creativity | Problem has decomposable structure | | Disruptive family (Disruptive) | assumption-destruction + lateral-thinking + perspective-forcing | Problem is stuck in dominant paradigm | | Coverage family (Coverage) | systematic-enumeration + morphological-exploration | Need exhaustive space mapping |
Comprehensive divergence: 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)
└── Campaign (10) — self-contained creative activity domain
└── Strategy (5 per campaign) — iterative framework with budget + state ledger
└── Tactic (2-3 per campaign) — SOP combination principle
└── SOP — 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 — load or create campaign context filecontext-checkpoint — 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 — no conclusions from snippets | | web-research | web-browsing | Must fetch full page via apify | | paper-overview | literature-engine | Abstract only — no methodology conclusions | | paper-search | literature-engine | AI summary — sufficient for methodology understanding | | paper-research | literature-engine | Full text — 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 |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | creative-ideation-paper-overview | Abstract-level paper scanning for broad coverage. Import of literature-engine/literature-overview skill. Abstract-level only — no methodology conclusions from abstracts. | | creative-ideation-paper-research | Deep paper analysis with full text reading. Import of literature-engine/literature-research skill. Full text access — required for quoting results, verifying claims, extracting detailed methodology. | | creative-ideation-paper-search | Mid-depth paper analysis via AI-generated summaries. Import of literature-engine/literature-search skill. Reads AI summary — sufficient for methodology understanding but not for quoting specific results. | | creative-ideation-web-research | Deep web page analysis with full content extraction. Import of web-browsing/web-research skill. Must fetch full page via apify — no shortcuts. | | creative-ideation-web-search | Quick web scanning for landscape understanding. Import of web-browsing/web-search skill. Snippets only — no conclusions from snippets alone. |
Optional, no fixed order; the final leaf is always a sop.
| Campaign | When to use | | --- | --- | | assumption-destruction | Assumption Destruction Campaign — open new solution spaces by negating, reversing, and challenging fundamental assumptions. | | biomimicry | Biomimicry Campaign — discover transferable solutions from biological systems via Design Spiral, BioTRIZ, functional analogy, ecosystem patterns, and evolution strategies. | | combinatorial-creativity | Combinatorial Creativity Campaign — produce emergent concepts via concept blending, multi-level bisociation, and function combination (Fauconnier-Turner) | | cross-domain-discovery | Cross-Domain Discovery Campaign — find transferable mechanisms from unrelated fields via bisociation, analogical transfer, random stimulus, and forced bridging | | lateral-thinking | Lateral Thinking Campaign — escape logical thinking tracks via PO/movement, random entry, concept fan, challenge, and six hats (de Bono) | | morphological-exploration | Morphological Exploration Campaign — systematic dimension-combination enumeration to discover unexplored solution spaces via Zwicky box, CCA, and GMA | | perspective-forcing | Perspective Forcing Campaign — discover hidden solutions by systematically switching viewpoints via roles, six hats, temporal projection, and constraint injection | | structural-deconstruction | Decompose systems into components and reassemble via SCAMPER, SIT, TRIZ, and recombination. Campaign orchestrating 5 strategies for systematic structural transformation. | | synectics | Synectics Campaign — systematic use of analogy and metaphor for breakthrough associations via Gordon's 4 analogy types and excursion method. | | systematic-enumeration | Systematic Enumeration Campaign — exhaustive coverage analysis to discover overlooked solution spaces via benchmark sweep, method-problem matrix, ablation, and failure taxonomy |
<!-- 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-01 | fail→pass | 16,928 | 29,741 | +76% | 1 | 1 | 0% | 2,647 | 7,193 | +172% | 0 | 0 | — |
case-02 | fail→pass | 21,993 | 37,842 | +72% | 1 | 1 | 0% | 3,539 | 8,271 | +134% | 0 | 0 | — |
case-03 | fail→pass | 6,556 | 7,340 | +12% | 1 | 1 | 0% | 997 | 3,191 | +220% | 0 | 0 | — |
case-04 | fail→pass | 30,230 | 5,533 | -82% | 1 | 1 | 0% | 5,049 | 2,924 | -42% | 0 | 0 | — |
case-05 | fail→pass | 27,362 | 11,621 | -58% | 1 | 1 | 0% | 4,946 | 4,027 | -19% | 0 | 0 | — |
case-18 | fail→pass | 17,314 | 2,201 | -87% | 1 | 1 | 0% | 763 | 2,335 | +206% | 0 | 0 | — |
case-06 | pass→pass | 12,432 | 7,091 | -43% | 1 | 1 | 0% | 1,841 | 3,004 | +63% | 0 | 0 | — |
case-07 | pass→pass | 8,888 | 4,227 | -52% | 1 | 1 | 0% | 1,308 | 2,565 | +96% | 0 | 0 | — |
case-08 | pass→pass | 11,267 | 4,087 | -64% | 1 | 1 | 0% | 1,812 | 2,605 | +44% | 0 | 0 | — |
case-09 | pass→pass | 11,535 | 5,256 | -54% | 1 | 1 | 0% | 1,708 | 2,669 | +56% | 0 | 0 | — |
case-10 | pass→pass | 12,826 | 5,365 | -58% | 1 | 1 | 0% | 1,971 | 2,816 | +43% | 0 | 0 | — |
case-11 | pass→pass | 21,611 | 3,997 | -82% | 1 | 1 | 0% | 3,609 | 2,603 | -28% | 0 | 0 | — |
case-12 | pass→pass | 13,710 | 4,470 | -67% | 1 | 1 | 0% | 2,212 | 2,682 | +21% | 0 | 0 | — |
case-13 | fail→pass | 13,312 | 12,200 | -8% | 1 | 1 | 0% | 2,105 | 2,642 | +26% | 0 | 0 | — |
case-14 | pass→pass | 14,053 | 5,847 | -58% | 1 | 1 | 0% | 2,424 | 2,897 | +20% | 0 | 0 | — |
case-15 | fail→pass | 18,535 | 3,786 | -80% | 1 | 1 | 0% | 2,702 | 2,632 | -3% | 0 | 0 | — |
case-16 | fail→pass | 28,168 | 3,929 | -86% | 1 | 1 | 0% | 925 | 2,637 | +185% | 0 | 0 | — |
case-17 | fail→pass | 13,327 | 3,430 | -74% | 1 | 1 | 0% | 2,217 | 2,510 | +13% | 0 | 0 | — |
case-19 | pass→pass | 14,117 | 2,441 | -83% | 1 | 1 | 0% | 1,966 | 2,310 | +17% | 0 | 0 | — |
case-20 | pass→pass | 10,363 | 2,882 | -72% | 1 | 1 | 0% | 1,540 | 2,407 | +56% | 0 | 0 | — |
case-21 | pass→pass | 14,563 | 2,656 | -82% | 1 | 1 | 0% | 2,215 | 2,448 | +11% | 0 | 0 | — |
case-22 | fail→pass | 9,040 | 2,401 | -73% | 1 | 1 | 0% | 1,213 | 2,333 | +92% | 0 | 0 | — |
case-23 | fail→fail | 26,915 | 210,060 | +680% | 1 | 1 | 0% | 4,015 | 8,179 | +104% | 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 21 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 +48 percentage points is the difference between those two pass rates over the 21 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.