/context-map

Create a detailed knowledge map of everything currently in context, showing what Claude knows and what it costs.

Context Knowledge Map

Use an agent to analyze and map all context components:

1. Files & Code Knowledge

For each file in context:

📄 src/main.py (8,500 tokens)
├─ Last accessed: 5 minutes ago
├─ Access count: 12 times
├─ Contains:
│  ├─ function main() - entry point
│  ├─ function process_data() - data processing
│  ├─ class DataHandler - data management
│  └─ 15 imports from external libraries
├─ Referenced by: src/utils.py, tests/test_main.py
├─ Depends on: config.json, utils.py
├─ Usage: ACTIVE (currently discussing)
└─ Prune recommendation: ❌ KEEP (active use)

2. Message History Knowledge

Analyze conversation context:

💬 Message History (45,000 tokens)
├─ Total messages: 87
├─ Time span: 2h 15m
├─ Topics covered:
│  ├─ Database schema design (20 messages, ~10K tokens)
│  ├─ Bug fixing in auth module (15 messages, ~8K tokens)
│  ├─ Code review discussions (12 messages, ~6K tokens)
│  ├─ General questions (40 messages, ~21K tokens)
├─ Compressible: ~15K tokens (33%)
│  └─ Old messages (>1h ago) could be summarized
└─ Prune recommendation: ⚠️  SUMMARIZE old messages

3. Tool Results Knowledge

Track tool output in context:

🔧 Tool Results (28,000 tokens)
├─ Read operations: 15 results (~12K tokens)
├─ Grep results: 8 results (~5K tokens)
├─ Bash outputs: 6 results (~8K tokens)
├─ Write confirmations: 10 results (~3K tokens)
├─ Oldest result: 45 minutes ago
├─ Still relevant: 8 results (~10K tokens)
└─ Prune recommendation: ✓ CLEAR 18K tokens of old results

4. System & Memory

System knowledge being tracked:

🧠 System Context (12,000 tokens)
├─ Session initialization
├─ Custom instructions
├─ Plugin configurations
├─ MCP server data
├─ Working directory info
└─ Prune recommendation: ❌ KEEP (required)

5. Knowledge Graph

Visualize relationships:

┌─────────────────────────────────────────────────┐
│ CONTEXT KNOWLEDGE GRAPH                         │
├─────────────────────────────────────────────────┤
│                                                 │
│  config.json ──────┬──────> main.py            │
│       ↓            │           ↓                │
│   schema.sql       └──────> utils.py            │
│       ↓                       ↓                 │
│   test-output.log        tests/test_*.py        │
│   [12K tokens]           [15K tokens]           │
│   ⚠️ PRUNE                ✓ KEEP                 │
│                                                 │
│  Message Flow:                                  │
│  [Initial Q] → [Exploration] → [Implementation] │
│  [40K tokens]   [25K tokens]    [15K tokens]    │
│  ⚠️ SUMMARIZE    ✓ KEEP         ✓ KEEP          │
└─────────────────────────────────────────────────┘

6. Token Allocation Breakdown

Detailed token accounting:

╔═══════════════════════════════════════════════╗
║ TOKEN ALLOCATION MAP                          ║
╠═══════════════════════════════════════════════╣
║                                               ║
║ Category          Tokens    %      Action    ║
║ ────────────────  ────────  ─────  ────────  ║
║ Active Files      35,000    24%    KEEP      ║
║ Inactive Files    18,000    12%    PRUNE     ║
║ Recent Messages   30,000    21%    KEEP      ║
║ Old Messages      15,000    10%    COMPRESS  ║
║ Fresh Tools       10,000     7%    KEEP      ║
║ Stale Tools       18,000    12%    CLEAR     ║
║ System Context    12,000     8%    KEEP      ║
║ Other              8,000     6%    REVIEW    ║
║ ────────────────  ────────  ─────  ────────  ║
║ TOTAL            146,000    100%             ║
║                                               ║
║ Prunable:         51,000    35%              ║
║ After Pruning:    95,000    65% → 47% of 200K║
║ Status Change:    🟡 Warning → 🟢 Healthy    ║
╚═══════════════════════════════════════════════╝

7. Knowledge Accessibility Score

Rate how useful current context is:

Knowledge Quality Metrics:
├─ Relevance Score: 78/100
│  └─ 22% of context not used in last 30min
├─ Efficiency Score: 65/100
│  └─ 35% prunable without information loss
├─ Coherence Score: 85/100
│  └─ Most context relates to current task
└─ Recommendation: PRUNE to increase efficiency

8. Smart Recommendations

Based on knowledge map:

📊 SMART RECOMMENDATIONS:

Immediate (High Value):
1. Remove test-output.log (12K tokens)
   - Last accessed: 45 mins ago
   - Content: Already reviewed
   - Risk: None (reproducible)

2. Clear stale Bash outputs (8K tokens)
   - Results from completed operations
   - No longer referenced
   - Risk: None

3. Prune schema.sql (18K tokens)
   - Last accessed: 60 mins ago
   - Original task completed
   - Risk: Low (can reload)

Medium Priority:
4. Summarize early messages (15K tokens → 5K)
   - Compress first 30 messages
   - Preserve key decisions
   - Risk: Low (detail loss acceptable)

Low Priority:
5. Remove backup-config.json (8K tokens)
   - Duplicate of main config
   - Never referenced
   - Risk: None

TOTAL SAVINGS: 51K tokens (35% reduction)
TIME TO EXECUTE: ~30 seconds

Agent-Accelerated Analysis

This command uses the Explore agent to:

• Rapidly scan all context components
• Build dependency graphs
• Calculate token costs accurately
• Identify optimization opportunities
• Generate actionable recommendations

Speed: ~10x faster than manual analysis Accuracy: Uses actual /context output + heuristics Coverage: Analyzes 100% of context automatically

Output Formats

Choose your preferred format:

• Visual (default): Rich ASCII tables and graphs
• JSON: Machine-readable for automation
• Markdown: Export for documentation
• Summary: Quick bullet points only

Let me generate your context knowledge map now using an agent for speed.