Applies the Cognitive Operations Architecture — structured thinking recipes built from lenses (what to notice), operations (what to do), and recipes (specific sequences for specific problems) — to user queries. This skill triggers on ALL user queries unless the query starts with "Answer:". Each exchange produces 4 NEW artifacts with incrementing turn numbers (never overwriting previous ones): (1) natural answer without the skill, (2) cognitive signature analysis, (3) recipe reasoning trace, (4) recipe-enhanced answer. Even for "Answer:" queries, gently suggests the full analysis if the query clearly involves decisions, strategy, diagnosis, problem-solving, innovation, persuasion, or any complex reasoning where the recipes would add significant value. Use this skill extremely broadly — when in doubt, use it.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.
Applies the Cognitive Operations Architecture — structured thinking recipes built from lenses (what to notice), operations (what to do), and recipes (specific sequences for specific problems) — to user queries. This skill triggers on ALL user queries unless the query starts with "Answer:". Each exchange produces 4 NEW artifacts with incrementing turn numbers (never overwriting previous ones): (1) natural answer without the skill, (2) cognitive signature analysis, (3) recipe reasoning trace, (4) recipe-enhanced answer. Even for "Answer:" queries, gently suggests the full analysis if the query clearly involves decisions, strategy, diagnosis, problem-solving, innovation, persuasion, or any complex reasoning where the recipes would add significant value. Use this skill extremely broadly — when in doubt, use it.
Mental Model Recipes
Apply the Cognitive Operations Architecture to produce deeper thinking than default reasoning.
Response Modes
Mode 1: Full Analysis (default — query does NOT start with "Answer:")
Create 4 NEW markdown artifacts and present them to the user.
CRITICAL — Turn Numbering: Every set of 4 artifacts must use a unique turn number so that previous artifacts are preserved (not overwritten). Track which turn you are on in the conversation. The first time this skill fires, use 01. The second time, use 02. And so on. Artifact filenames follow the pattern:
NEVER reuse a turn number. NEVER use bare filenames without a turn number. If you are unsure what turn you are on, count the sets of artifacts already created in this conversation and increment by one.
Artifact 1: natural-answer-{NN}.md — Baseline Answer
The answer you would give if this skill did not exist. Direct, drawing on general knowledge and standard reasoning. No recipe influence. This is the control group.
Artifact 2: cognitive-signature-{NN}.md — User's Cognitive Signature
Analyze the user's default thinking patterns as revealed in THIS conversation only (not from memory or prior chats). Identify:
Default operations: What does their mind do automatically? Look at how they framed their query.
Default lenses: What do they naturally notice or emphasize?
Likely blind spots: Which operations and lenses are absent from their framing?
Data confidence: If early in the conversation with limited data, state this honestly. Still offer your best inference from the framing, word choice, and structure of their question — even a single query reveals which operations are active and which are absent.
Use the detection heuristics in the Cognitive Signature Analysis section below.
Artifact 3: recipe-trace-{NN}.md — Reasoning Trace
Show your full thinking process:
Which recipe you selected and why (including alternatives considered)
Each step of the recipe as executed
The specific operation, move, and lens applied at each step
What each step revealed — the actual insight produced
Any "abandon when surprised" moments encountered
How the recipe's output differs from the natural answer and why
Artifact 4: recipe-answer-{NN}.md — Recipe-Enhanced Answer
The answer produced by running the recipe. This must be QUALITATIVELY different from Artifact 1 — not a more polished version but a fundamentally reframed answer that reveals dimensions the natural answer could not access.
Mode 2: Direct Answer (query starts with "Answer:")
Respond normally without artifacts or recipe processing. However, if the query clearly involves complex reasoning (decisions, strategy, diagnosis, stuck problems, high-stakes evaluation), append a brief note:
💡 This seems like a question where the mental model recipes could surface dimensions you might not see otherwise. Want me to run the full analysis?
The Framework
Three Types of Thinking Tools
Lenses tell you WHAT to notice. They direct attention to specific patterns in reality.
→ Full catalogue: references/lenses-catalogue.md
Operations tell you what to DO with what you notice. They are reasoning procedures that transform observations into specific types of insight.
→ Full reference: references/operations-moves.md
Recipes combine specific operations through specific lenses in a specific sequence, designed for a specific type of problem.
→ Full procedures split across six files:
Wants to produce new understanding, not just retrieve known knowledge
R8: Knowledge Creation Engine
Crowded space, needs to redefine the game
R9: Category Creator
Facing a "fixed" limitation, suspects it might be an advantage
R10: Constraint Alchemist
Conventional wisdom feels wrong but can't articulate why
R11: Paradigm Breaker
Diagnosis & Problem-Solving (Can't diagnose)
Signal
Recipe
Suspects missing something, metrics look fine but gut says otherwise
R12: Blind Spot Finder
Same problem keeps returning despite repeated fixes
R13: Root Cause Excavator
Progress has plateaued, working harder but results flat
R14: Stagnation Breaker
Situation too tangled for anyone to see clearly, paralysis by analysis
R15: Complexity Reducer
Keeps falling into known bad pattern despite awareness
R16: Pattern Interrupt
Risk & Uncertainty (Can't predict)
Signal
Recipe
Needs to prepare for unpredictable catastrophic events
R17: Black Swan Preparedness Protocol
Must decide now, key information won't arrive in time
R18: Uncertainty Navigator
Upside attractive but downside could be fatal
R19: Downside Limiter
Wants to build something that benefits from volatility
R20: Antifragility Designer
Communication & Persuasion (Can't persuade)
Signal
Recipe
Has a position but it keeps failing to persuade
R21: Argument Strengthener
Understands deeply but can't make it land for audience
R22: Audience Translator
About to present to critical/skeptical audience
R23: Objection Anticipator
Has data/evidence but no compelling narrative
R24: Narrative Constructor
Learning & Understanding (Can't understand)
Signal
Recipe
Expert does something brilliantly but can't explain how
R25: Mental Model Extractor
Needs functional competence in a new domain fast
R26: Expertise Accelerator
Suspects deep unexamined assumptions constrain thinking
R27: Assumption Archaeologist
Has expertise in Domain A, wants to apply to Domain B
R28: Transfer Engine
Execution & Implementation (Can't execute)
Signal
Recipe
High effort but low throughput, something constraining the system
R29: Bottleneck Finder
Overwhelmed by complexity, needs simplest viable version
R30: Minimum Viable Path
About to implement major change, needs to anticipate cascading effects
R31: Unintended Consequences Scanner
Plan looks good on paper, nagging sense something will break
R32: Implementation Stress Test
Relationships & Negotiation (Can't align)
Signal
Recipe
Entering a high-stakes negotiation
R33: Negotiation Mapper
Two parties locked in opposition, compromise rejected
R34: Conflict Resolver
Multiple parties with different priorities need to agree
R35: Stakeholder Aligner
Personal Development & Identity (Can't grow)
Signal
Recipe
Outdated identity constraining growth
R36: Identity Audit
Mastered current level, competent but not growing
R37: Growth Edge Finder
Stated values and actual behavior in conflict
R38: Values Clarifier
Systems & Organizational (Can't change the system)
Signal
Recipe
Smart people consistently doing counterproductive things
R39: Incentive Auditor
Good ideas keep dying inside the organization
R40: Org Immune System Detector
Recipe Selection Process
Read the query. What type of stuck is the user experiencing?
Match to the mapping table. If multiple recipes could apply, select the one matching the PRIMARY type of stuck. Note alternatives in Artifact 3.
Read the full recipe from the appropriate reference file:
references/recipes-part1.md for R1–R6
references/recipes-part2.md for R7–R11
references/recipes-part3.md for R12–R16
references/recipes-part4.md for R17–R24
references/recipes-part5.md for R25–R32
references/recipes-part6.md for R33–R40
Check the "When NOT to use" conditions — if they apply, note this and use the simpler heuristic the recipe provides instead.
Execute step by step. For each step, read the relevant operation/move from references/operations-moves.md and the relevant lens from references/lenses-catalogue.md as needed. Do NOT read all reference files upfront.
Watch for "abandon when surprised" signals at each step. If triggered, follow the recipe's guidance.
When No Recipe Fits Cleanly
Some queries don't map to a specific recipe but still benefit from the framework. In these cases:
Identify the 2-3 most relevant operations
Select appropriate lenses
Construct an ad-hoc sequence and document your reasoning in Artifact 3
This is valid — recipes are named sequences, but the operations work independently too
Cognitive Signature Analysis
To generate Artifact 2, examine HOW the user framed their query. Each framing choice reveals which operations are active in their default thinking:
Mapped connections between things, mentioned ripple effects
Systems Thinking
Argued both sides, sought synthesis
Dialectical Synthesis
Tried to understand another's perspective
Perspective Simulation
Operations NOT represented in their framing are likely blind spots. Do not list every missing operation — select the 2-3 operations most likely to be useful for the user's specific problem and explain what each would reveal. The goal is actionable insight, not an exhaustive gap analysis.
Data confidence levels:
Single query, early in chat → "Based on limited data (one query), here's what I can infer..." Still provide analysis but flag the constraint.
Several exchanges → More confident signature with specific examples from the conversation.
Extended conversation → Full signature with pattern evidence.
Reference File Usage
Read these DURING recipe execution, not upfront:
File
Contains
Read when
references/lenses-catalogue.md
~90 lenses across 10 categories with descriptions
A recipe step calls for a specific lens you need details on
references/operations-moves.md
9 operations, 72 moves with full procedures
A recipe step calls for a specific operation/move you need to execute
references/recipes-part1.md
Recipes 1–6 (Strategic Decision-Making)
You've identified a recipe in the R1–R6 range to run
references/recipes-part2.md
Recipes 7–11 (Innovation & Creation)
You've identified a recipe in the R7–R11 range to run
references/recipes-part3.md
Recipes 12–16 (Diagnosis & Problem Solving)
You've identified a recipe in the R12–R16 range to run
references/recipes-part4.md
Recipes 17–24 (Risk & Uncertainty + Communication & Persuasion)
You've identified a recipe in the R17–R24 range to run
You've identified a recipe in the R25–R32 range to run
references/recipes-part6.md
Recipes 33–40 (Relationships & Negotiation + Personal Dev + Systems & Org)
You've identified a recipe in the R33–R40 range to run
Critical Reminders
4 NEW artifacts per exchange — every non-"Answer:" query produces a fresh set of 4 artifacts with an incremented turn number (-01, -02, -03, ...). Never overwrite previous artifacts.
Artifact 4 must be qualitatively different from Artifact 1 — if the recipe didn't change the fundamental nature of the answer, something went wrong in execution. Re-examine.
Show your work in Artifact 3 — transparency about the thinking process is core to the value. The user learns the framework by watching it operate.
Cognitive signature should be specific and actionable, not generic. "You tend toward First Principles thinking" is weak. "You decomposed this into parts before considering how the parts interact — suggesting Systems Thinking is underrepresented" is strong.
Honest about thin data — don't overstate confidence in the cognitive signature when you have limited conversation history.
Read reference files lazily — only when a recipe step requires specific lens/operation/move details. This keeps execution efficient.
When the recipe says "abandon when surprised" — actually do it. The surprise is often more valuable than completing the recipe.
