Instruction file imported from fabioc-aloha/ai-wallpaper-generator (.github/instructions/deep-thinking.instructions.md). Copyright stays with the author.

Deep Thinking Cognitive Skill

Activation Triggers

Explicit Triggers (User Request)

• "think deep" / "think deeply"
• "deep analysis"
• "analyze thoroughly"
• "systematic thinking"
• "reason through this"

Implicit Triggers (Self-Activation)

• When something isn't working: Approach fails, unexpected results, repeated errors
• Important decisions: Architecture choices, irreversible changes, trade-off evaluations
• Uncertainty detected: Multiple viable paths, conflicting evidence, knowledge gaps
• High-stakes situations: Production systems, security concerns, data integrity

Deep Thinking Protocol

Phase 1: Problem Domain Identification

Objective: Establish clear boundaries and context for the problem

Actions:

1. Classify domain type: Technical, conceptual, interpersonal, strategic, creative
2. Identify key constraints: Time, resources, dependencies, unknowns
3. Map stakeholders: Who is affected, who decides, who implements
4. Define success criteria: What does "solved" look like?

Output: Domain classification with constraint matrix

Phase 2: Episodic Memory Scan

Objective: Retrieve relevant past experiences and learned patterns

Actions:

1. Check Skill Selection Optimization results — if SSO ran (complex task), use its skill list as search guide
2. Scan episodic memory (.prompt.md files) for similar situations
3. Search skills (skills/*/SKILL.md files) for applicable frameworks
4. Review meditation sessions (.github/episodic/) for consolidated insights
5. Check procedural memory (.instructions.md files) for established protocols

Search Patterns:

• Direct keyword matching on problem domain
• Analogical reasoning for structurally similar problems
• Cross-domain pattern recognition

Output: Ranked list of relevant episodes with similarity scores

Phase 3: Learning Extraction

Objective: Distill actionable insights from retrieved memories

Actions:

1. Extract success patterns: What worked in similar situations?
2. Identify failure modes: What pitfalls to avoid?
3. Note contextual factors: What conditions enabled success?
4. Synthesize meta-patterns: What principles emerge across episodes?

Learning Categories:

Output: Structured learning extraction with categorization

Phase 4: Pattern Application

Objective: Apply extracted learnings to current problem

Actions:

1. Map patterns to problem: Align extracted insights with current constraints
2. Adapt for context: Modify approaches based on current conditions
3. Synthesize solution approach: Combine applicable patterns
4. Identify gaps: Note where past experience doesn't fully apply
5. Generate recommendations: Concrete, actionable steps

Application Framework:

Current Problem + Extracted Patterns + Contextual Adaptation = Solution Approach

Output: Synthesized solution approach with adaptation notes

Phase 5: Source Citation

Objective: Provide transparency and enable verification

Citation Format:

**Source Episodes**:
- [episode-name.prompt.md](path) - "Key insight extracted"
- [relevant-skill/SKILL.md](path) - "Applicable framework"
- [meditation-session-date.prompt.md](path) - "Consolidated learning"

Citation Requirements:

• Link to specific memory file
• Quote or summarize the relevant insight
• Note confidence level (high/medium/low)
• Indicate adaptation required for current context

Integration with Meta-Cognitive Awareness

Before Deep Thinking:

• Acknowledge activation: "Engaging deep thinking protocol..."
• State the problem as understood

During Deep Thinking:

• Progress indicators for each phase
• Surface key insights as they emerge

After Deep Thinking:

• Summarize approach with confidence assessment
• Identify areas requiring further investigation
• Offer to explore specific aspects in more depth

Quality Assurance

Verification Checklist

• Problem domain clearly identified
• Episodic memory thoroughly scanned
• Learnings explicitly extracted (not assumed)
• Patterns adapted to current context
• Sources properly cited
• Gaps and uncertainties acknowledged

Anti-Patterns to Avoid

• Premature solution: Jumping to answers without phase completion
• Memory fabrication: Citing non-existent episodes
• Over-generalization: Applying patterns without contextual adaptation
• Citation omission: Presenting insights without source attribution

Synapses

High-Strength Bidirectional Connections

• [.github/instructions/alex-core.instructions.md] (Critical, Extends, Bidirectional) - "Meta-cognitive deep reasoning enhancement"
• [.github/instructions/bootstrap-learning.instructions.md] (High, Integrates, Bidirectional) - "Learning extraction and application protocols"
• [.github/instructions/skill-selection-optimization.instructions.md] (High, Coordinates, Bidirectional) - "SSO provides skill survey results; Deep Thinking uses them for episodic scan"

Medium-Strength Output Connections

• [.github/instructions/embedded-synapse.instructions.md] (High, Leverages, Forward) - "Pattern recognition across memory network"
• [.github/instructions/worldview-integration.instructions.md] (Medium, Validates, Forward) - "Ethical reasoning during analysis"
• [.github/prompts/cross-domain-transfer.prompt.md] (High, Enables, Forward) - "Cross-domain pattern application"

Input Connections

• [.github/prompts/unified-meditation-protocols.prompt.md] (Medium, Receives, Backward) - "Consolidated insights for retrieval"
• [.github/prompts/domain-learning.prompt.md] (Medium, Receives, Backward) - "Acquired domain knowledge integration"

Primary Function: Execute systematic deep thinking with episodic memory integration for comprehensive problem analysis.

Activation Triggers:

• Complex problem requiring multi-phase analysis
• User requests deep or thorough thinking
• Situations benefiting from past experience application
• Problems requiring cited, verifiable reasoning

Output Expectations:

• Structured analysis following 5-phase protocol
• Explicit citations to source episodes
• Confidence levels on recommendations
• Identified gaps and uncertainties