[P3] Implement Knowledge Gardening Agent

[Dan-StrategicAutomation]

Description

Automated maintenance and curation of the Knowledge Base to keep it clean and relevant.

Current Issues

• No automatic cleanup of duplicate learnings
• Overly specific learnings never get generalized
• Old/outdated learnings accumulate
• AI.md can become verbose

Features

• Duplicate Detection: Find and merge similar learnings
• Pattern Generalization: Convert specific examples to general patterns
• Archival: Move learnings older than threshold to archive
• AI.md Optimization: Regenerate for clarity and conciseness
• Quality Scoring: Identify low-quality learnings for review

Implementation

Create KnowledgeGardener agent that:

1. Analyzes all learnings
2. Identifies improvements
3. Proposes changes
4. (Optional) Auto-applies with confidence threshold

Can run:

• Manually: uv run python cli.py garden-knowledge
• Scheduled: Weekly via cron/GitHub Actions
• Triggered: When KB reaches size threshold

Tasks

• Design gardening algorithm
• Implement duplicate detection (similarity threshold)
• Implement pattern generalization agent
• Add archival logic (date-based)
• Add quality scoring
• Create CLI command
• Add configuration options
• Update documentation

Impact

🌱 Keeps KB healthy and prevents degradation over time.

Files to Create

• agents/knowledge_gardener.py
• workflows/garden_knowledge.py

Priority

Low - Useful but not urgent for small KBs