system-learn

Ingest new procedural memory (skills, patterns, docs) into the vector database.

⭐88stars🔀9forks📁View on GitHub🕐Updated Dec 30, 2025

When & Why to Use This Skill

The system-learn skill enables Claude agents to build a persistent procedural memory by ingesting markdown-based skills, patterns, and documentation into a vector database (ChromaDB). This allows the agent to 'learn' from past experiences and technical guides, ensuring high-quality semantic retrieval and long-term context retention for complex, repetitive tasks.

Use Cases

  • 1. Technical Knowledge Base Indexing: Ingesting new API documentation or internal technical guides to ensure the agent can accurately reference them in future coding sessions.
  • 2. Coding Pattern Standardization: Saving successful architectural patterns or reusable code snippets identified during development to maintain consistency across a large codebase.
  • 3. Workflow Automation: Learning and storing Standard Operating Procedures (SOPs) from markdown files to help the agent follow specific organizational protocols automatically.
namesystem-learn
descriptionIngest new procedural memory (skills, patterns, docs) into the vector database.

Learning Skill

This skill allows the agent to "learn" new patterns, skills, or documentation by ingesting markdown files into the procedural memory collection in ChromaDB.

Usage

Use this when you have identified a reusable pattern, wrote a new guide, or want to index a documentation file for future semantic retrieval.

uv run python .fleet/context/scripts/memory_manager.py learn --file <path_to_markdown_file>

Arguments

  • --file: The absolute or relative path to the markdown file you want to learn.

Best Practices

  • Ensure the markdown file has a clear title (# Title).
  • The file should contain reusable information, not just ephemeral session data.
  • Save the file in .fleet/context/skills/ before learning it to keep the source organized.