Context Aggregator

Context Aggregator

Generate unified session summaries that consolidate conversation history, file operations, and resource usage into comprehensive reports. This skill is essential for job completion notifications, progress reports, and audit trails.

When to Use

• Job completion: Generate a final summary when finishing a job
• Progress updates: Provide mid-session summaries to the user
• Handoff reports: Document what was accomplished for human review
• Audit trails: Create complete records of agent activity
• Cost reporting: Track token usage and API costs across sessions

How It Works

The aggregator pulls from multiple data sources:

1. Session logs: Conversation history (user/assistant messages)
2. File operations: Reads, writes, and edits performed
3. Token usage: LLM token consumption and cost tracking

Usage

Generate a full session summary

context-aggregator summary

Generate a brief summary (for notifications)

context-aggregator brief

Export as JSON (for programmatic use)

context-aggregator json

Show only conversation highlights

context-aggregator conversation

Show only file changes

context-aggregator files

Show cost breakdown

context-aggregator cost

Generate a complete report

context-aggregator report --output /job/logs/session-report.md

Include file diffs in report

context-aggregator report --include-diffs --output /job/logs/session-report.md

Output Formats

Brief Summary (for notifications)

📊 Session Summary
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
⏱️  Duration: 23 minutes
💬 Messages: 12 (4 user, 8 assistant)
📁 Files: 15 operations (3 reads, 8 edits, 4 writes)
🔢 Tokens: 45,230 (~$0.82)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🎯 Accomplished:
• Created data processing pipeline
• Fixed authentication bug
• Updated API documentation

Full Report (Markdown)

# Session Report

**Generated:** 2026-02-25 14:30:00 UTC
**Session ID:** job-abc123

## Conversation Summary

| Time | Role | Summary |
|------|------|---------|
| 14:07 | User | Analyze the codebase for performance issues |
| 14:08 | Assistant | I'll analyze the codebase using the code-indexer... |
| 14:15 | Assistant | Found 3 critical bottlenecks in the processing pipeline |
| 14:20 | User | Fix the main bottleneck |
| 14:28 | Assistant | Implemented caching solution, 10x speedup achieved |

**Total Messages:** 12 (4 user, 8 assistant)

## File Operations

### Reads (3)
- `/job/src/main.ts` - Loaded entry point
- `/job/src/processor.ts` - Analyzed processing logic
- `/job/config/settings.json` - Read configuration

### Edits (8)
- `/job/src/cache.ts` - Added LRU cache implementation
- `/job/src/processor.ts` - Integrated cache, optimized loops
- `/job/tests/processor.test.ts` - Updated tests

### Writes (4)
- `/job/logs/analysis.md` - Performance analysis report
- `/job/logs/optimization.md` - Optimization recommendations

## Resource Usage

| Metric | Value |
|--------|-------|
| Total Tokens | 45,230 |
| Input Tokens | 28,450 |
| Output Tokens | 16,780 |
| Estimated Cost | $0.82 |
| Duration | 23 minutes |

## Key Accomplishments

1. **Performance Analysis**: Identified 3 critical bottlenecks
2. **Cache Implementation**: Added LRU cache reducing DB queries
3. **Test Coverage**: Updated tests with new test cases
4. **Documentation**: Created optimization recommendations

JSON Output

{
  "sessionId": "job-abc123",
  "generatedAt": "2026-02-25T14:30:00Z",
  "duration": "23 minutes",
  "conversation": {
    "totalMessages": 12,
    "userMessages": 4,
    "assistantMessages": 8,
    "highlights": [
      "User requested performance analysis",
      "Identified 3 critical bottlenecks",
      "Implemented LRU cache solution"
    ]
  },
  "files": {
    "total": 15,
    "reads": 3,
    "edits": 8,
    "writes": 4,
    "paths": [
      "/job/src/main.ts",
      "/job/src/processor.ts",
      "/job/src/cache.ts"
    ]
  },
  "usage": {
    "totalTokens": 45230,
    "inputTokens": 28450,
    "outputTokens": 16780,
    "estimatedCost": 0.82,
    "currency": "USD"
  }
}

Integration with Other Skills

• With email-agent: Send session summary via email on job completion
• With voice-output: Announce summary when job completes
• With memory-agent: Store session summary for future reference
• With session-files: Provides file operation data
• With model-usage: Provides token/cost data

Session Data Locations

The aggregator looks for data in these locations:

• Conversation: /job/logs/<job-id>/session.jsonl
• File operations: /job/logs/<job-id>/files.jsonl
• Token usage: /job/logs/<job-id>/usage.jsonl

Tips

1. Run early and often: Use context-aggregator brief during long jobs for checkpoints
2. Custom output: Use --format markdown or --format json for different needs
3. Filter concerns: Use --include-files or --include-conversation to focus output
4. Automated reporting: Add to job completion workflow for automatic reporting