cytostack/openwolf

• **Archetype**: claude-code-enhancement-middleware
• **Stars**: 467
• **Last pushed**: 2026 (recent)
• **License**: AGPL-3.0
• **Discovered**: 2026-04-13
• **Skills found**: 0 (middleware tool, not skill collection)

Summary

Middleware that enhances Claude Code with project intelligence, token tracking, and learning memory through 6 hook scripts. Features file indexing via anatomy.md with token estimates, learning memory through cerebrum.md capturing preferences and corrections, token ledger tracking lifetime usage across 132+ sessions with 65.8% average reduction, and bug memory preventing rediscovery of known issues. Autonomous middleware that manages context without workflow changes.

Assessment

HIGH VALUE for babysitter enhancement patterns. The learning memory system, context-aware caching layer, and autonomous middleware approach provide excellent models for babysitter improvements. The anatomy.md file indexing, cerebrum.md preference capture, and buglog.json issue tracking demonstrate sophisticated context management. The 65.8% token reduction across 132+ sessions with lifecycle hook integration shows proven effectiveness.

Extraction Priority

• High
• Rationale: The learning memory patterns, context-aware caching, and autonomous middleware architecture are directly applicable to babysitter's orchestration model. The file indexing, preference tracking, and bug memory systems provide sophisticated context management that could enhance babysitter's run persistence and retrospective capabilities.

Processes

• **Learning Memory System**: Capture user preferences, corrections, and known mistakes across sessions
• **Context-Aware File Indexing**: Generate project anatomy with token estimates to optimize file access
• **Token Usage Analytics**: Track and optimize token consumption across sessions with detailed metrics
• **Bug Memory Prevention**: Maintain issue history to prevent rediscovery of known problems
• **Autonomous Context Management**: Manage project intelligence without requiring workflow changes
• **Design QC Process**: Screenshot-based UI evaluation and quality assessment

Plugin Ideas

• **OpenWolf Integration**: Direct integration with OpenWolf middleware for enhanced babysitter context management

Patterns

• Context-aware caching layer with usage-based learning
• Anatomy-based file indexing with token cost estimation
• Cerebrum learning memory for preference persistence
• Token ledger with session-based analytics
• Hook-based lifecycle interception
• Bug memory with issue prevention
• Reframe framework migration assistance
• Design QC with screenshot evaluation

Library Mapping

Plugin Marketplace Mapping