Context Compressor Skill

<identity> Context Compressor Skill - Techniques for reducing token usage while preserving decision-critical information. Helps agents work efficiently within context limits. </identity> <capabilities> - Compressing conversation history - Summarizing code and documentation - Extracting key decisions and context - Creating efficient memory snapshots - Reducing redundancy in context </capabilities> <instructions> <execution_process>

Step 1: Identify Compressible Content

Content types that can be compressed:

Type	Compression Strategy
Code	Keep signatures, summarize implementations
Conversations	Extract decisions, drop small talk
Documentation	Keep headings and key points
Errors	Keep message and location, drop stack frames
Logs	Keep patterns, drop repetitions

Step 2: Apply Compression Techniques

Technique 1: Decision Extraction

Before:

User: Should we use Redis or Memcached?
Assistant: Let me analyze both options...
[500 words of analysis]
Recommendation: Redis for pub/sub support.
User: Ok let's use Redis.

After:

Decision: Use Redis (chosen for pub/sub support)

Technique 2: Code Summarization

Before:

// 100 lines of UserService implementation

After:

UserService: CRUD operations for users
- Methods: create, read, update, delete, findByEmail
- Dependencies: db, validator, logger
- Location: src/services/user.js

Technique 3: Error Compression

Before:

Error: Cannot read property 'id' of undefined
    at UserController.getUser (src/controllers/user.js:45:23)
    at Layer.handle [as handle_request] (node_modules/express/lib/router/layer.js:95:5)
    ... 20 more stack frames

After:

Error: Cannot read 'id' of undefined @ src/controllers/user.js:45
Cause: User object is null when accessing .id

Step 3: Structure Compressed Output

Use consistent formats:

## Session Summary

### Decisions Made

- [D1] Use Redis for caching
- [D2] JWT for authentication

### Files Modified

- src/auth/jwt.js (new)
- src/config/redis.js (updated)

### Open Items

- [ ] Add rate limiting
- [ ] Write tests for JWT

Step 4: Validate Compression

Ensure critical info preserved:

• All decisions captured
• Key file locations retained
• Error causes documented
• Next steps clear

</execution_process>

<best_practices>

1. Preserve Decisions: Never lose decision rationale
2. Keep Locations: File paths and line numbers are critical
3. Summarize, Don't Delete: Transform verbose content
4. Use References: Point to files instead of including content
5. Test Recovery: Can you continue work from compressed context?

</best_practices> </instructions>

<examples> <usage_example> **Compression Request**:

Compress the last 10 messages for context handoff

Example Response:

## Context Summary (compressed from 10 messages)

### Task

Implementing user authentication with JWT

### Progress

1. Created JWT utility (src/auth/jwt.js)
2. Added login endpoint (src/routes/auth.js)
3. Tests failing - mock issue

### Key Decisions

- Using RS256 for JWT signing (security over performance)
- 1-hour token expiry with refresh tokens

### Current Issue

Jest mock for jwt.sign not working
Location: src/auth/**tests**/jwt.test.js:23

### Next Steps

1. Fix JWT mock
2. Add refresh token endpoint

</usage_example> </examples>

Rules

• Never lose decision rationale
• Always include file locations
• Test that work can continue from compressed context

Workflow Integration

This skill supports multi-agent orchestration by enabling efficient context management:

Router Decision:

.claude/workflows/core/router-decision.md

• Router spawns agents that use this skill for context-efficient handoffs
• Used in long-running sessions to maintain continuity

Artifact Lifecycle:

.claude/workflows/core/skill-lifecycle.md

• Compression patterns evolve with framework changes
• Session summaries feed into memory protocol

Related Workflows:

• session-handoff

  skill for complete handoff protocol

• swarm-coordination

  skill for multi-agent context sharing

---

Memory Protocol (MANDATORY)

Before starting:

cat .claude/context/memory/learnings.md

After completing:

• New pattern ->

  .claude/context/memory/learnings.md

• Issue found ->

  .claude/context/memory/issues.md

• Decision made ->

  .claude/context/memory/decisions.md

ASSUME INTERRUPTION: Your context may reset. If it's not in memory, it didn't happen.