The Agent Skills Directory

[SAFE]: The skill is designed to manage documentation within the .lattice/ directory. It reads and writes to local markdown files such as .lattice/learnings/operational-learnings.md based on user-confirmed project patterns.
[SAFE]: Human-in-the-loop design: The 'Harvest Behavior' and 'Tighten Behavior' sections explicitly state that the agent never writes autonomously and must obtain user approval for every entry or modification.
[SAFE]: No network operations, remote dependencies, or obfuscated content are present. The skill relies entirely on natural language instructions for synthesizing session data into project documentation.
[SAFE]: Minimal privilege: The skill defines clear scope boundaries and resolution logic for its configuration, avoiding access to sensitive system directories or credentials.

learning-harvest