/contexte

⚡ EXECUTION INSTRUCTIONS FOR CLAUDE

When this command is invoked, YOU (Claude) must execute these steps immediately: This is NOT documentation - these are COMMANDS to execute right now. Use TodoWrite to track progress through multi-phase workflows.

🚨 EXECUTION WORKFLOW

🚨 THREE-PHASE EXECUTION WORKFLOW

Action Steps:

1. Review the reference documentation below and execute the detailed steps.

📊 PHASE 1: CONTEXT ESTIMATION (UNIVERSAL COMPOSITION)

Action Steps: MANDATORY: Execute Built-in /context Command First

1. 🎯 CALL /context COMMAND DIRECTLY - Use Claude Code's built-in context analysis
2. Extract actual token metrics from /context output (not estimated)
3. Parse real context breakdown (Messages, MCP tools, Memory files, etc.)
4. Use actual percentages and token counts from /context command output
5. Never estimate or guess context usage - only use /context command data

🔍 PHASE 2: STRATEGIC ANALYSIS

Action Steps: Pattern Recognition & Optimization Detection

1. Context breakdown by operation type (reads, searches, tool calls)
2. Identify context-heavy operations and patterns in current session
3. Analyze file read efficiency and sizes for optimization opportunities
4. Evaluate API response complexity and tool usage patterns

💡 PHASE 3: ACTIONABLE RECOMMENDATIONS

Action Steps: Tailored Optimization Guidance

1. Specific optimization suggestions tailored to current session state
2. Serena MCP integration opportunities for efficiency gains
3. Context-efficient workflow alternatives for detected patterns
4. Strategic checkpoint and recovery recommendations based on usage

Phase 5: 🎯 MANDATORY EXECUTION SEQUENCE:

Action Steps:

1. Review the reference documentation below and execute the detailed steps.

⚡ PHASE 1 EXECUTION (UNIVERSAL COMPOSITION)

Action Steps: MANDATORY: Execute /context Command First

1. 🎯 EXECUTE /context COMMAND - Get actual context data from Claude Code CLI
2. Parse /context output - Extract real token counts, percentages, breakdowns
3. Use actual metrics only - Never estimate, always use /context command data
4. Display actual context status - Show real usage from /context output
5. Determine health status based on actual percentage from /context

🔍 PHASE 2 EXECUTION

Action Steps: Strategic Analysis Based on Phase 1 Results

1. Analyze Phase 1 metrics for optimization opportunities
2. Identify context-heavy operations (large file reads, repeated searches)
3. Detect inefficient tool usage patterns from session data
4. Evaluate potential for Serena MCP optimization
5. Assess checkpoint timing recommendations

💡 PHASE 3 EXECUTION

Action Steps: Actionable Recommendations Delivery

1. Provide context-efficient alternatives for detected patterns
2. Deliver strategic workflow improvements based on analysis
3. Offer tool selection hierarchy guidance
4. Present session management strategies
5. Conclude with immediate next steps

Phase 9: 🚨 EXECUTION INSTRUCTIONS FOR CLAUDE

Action Steps: When /contexte is invoked, MUST EXECUTE ALL THREE PHASES SEQUENTIALLY:

📊 PHASE 1: CONTEXT ESTIMATION (MANDATORY FIRST) - UNIVERSAL COMPOSITION

Action Steps:

🎯 EXECUTE BUILT-IN `/context` COMMAND:
1. ALWAYS call `/context` command first to get actual data
2. Extract real token counts from `/context` output
3. Parse actual breakdown (Messages, MCP tools, Memory files, etc.)
4. Use real percentages and usage metrics from `/context`
5. Never estimate or calculate - only use `/context` command data
6. Display actual context status based on `/context` output

🔍 PHASE 2: STRATEGIC ANALYSIS (BASED ON PHASE 1)

Action Steps:

Analyze Phase 1 results for optimization opportunities:
1. Identify specific context-heavy operations from session
2. Suggest context-efficient alternatives for current patterns
3. Recommend Serena MCP opportunities where applicable
4. Provide strategic checkpoint guidance based on usage
5. Offer workflow improvement suggestions tailored to session

💡 PHASE 3: ACTIONABLE RECOMMENDATIONS (FINAL PHASE)

Action Steps:

Deliver immediate actionable advice:
1. Specific optimization actions for current session
2. Tool selection hierarchy improvements
3. Session management strategies
4. Future workflow enhancements
5. Next steps for context efficiency

📋 REFERENCE DOCUMENTATION

Context Usage Estimation Command

Usage: /contexte or /con

Purpose: First run context estimation, then provide comprehensive analysis with optimization recommendations for Claude Code CLI conversations.

🎯 UNIVERSAL COMPOSITION PRINCIPLE

CRITICAL: This command uses Universal Composition - it MUST call the built-in /context command first to get actual data, then provide strategic analysis based on that real data.

❌ NEVER estimate context usage
✅ ALWAYS call /context command first
✅ ALWAYS use actual data from /context output

Implementation

Execution Method: Three-phase sequential analysis workflow

Context Estimation Algorithm:

1. Tool Usage Analysis: Count and categorize all tool operations
2. Content Size Estimation: Approximate tokens from tool outputs and responses
3. Complexity Scoring: Weight different operation types by context consumption
4. Optimization Detection: Identify inefficient patterns and suggest improvements

Token Estimation Rules:

• Base conversation: ~500-1000 tokens
• Tool operations: 100-500 tokens each (varies by type)
• File reads: Estimated by file size (chars ÷ 4)
• Web searches: ~200-800 tokens per search
• Large responses: Actual character count ÷ 4
• Serena MCP: 50-200 tokens (very efficient)

Claude Sonnet 4 Limits:

• Enterprise: 500K tokens
• Paid Plans: 200K tokens
• Estimation Accuracy: ±20% variance expected

Output Formats

Basic Output:

📊 CONTEXT USAGE ESTIMATION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🔹 Estimated Tokens Used: ~15,400 / 500,000 (3.1%)
🔹 Session Complexity: Medium (Score: 34/100)  
🔹 Tools Used: 8 operations (4 types)
🔹 Context Status: ✅ HEALTHY

💡 Quick Tip: Consider /checkpoint if planning complex analysis

Detailed Analysis:

• Token breakdown by operation type
• Identification of context-heavy operations
• File read patterns and efficiency analysis
• Optimization opportunities and recommendations

Integration

Auto-triggered Recommendations:

• High complexity sessions (60+ score): Suggest optimization
• Large file operations detected: Recommend Serena MCP
• Context approaching 50%: Recommend checkpoint

Command Usage:

/contexte    # Execute all three phases: Context Estimation → Strategic Analysis → Actionable Recommendations
/con         # Alias for /contexte