The Agent Skills Directory

[SAFE]: Automatic Learning Triggers
The plugin monitors session turn counts and elapsed time to trigger memory updates. This is a functional feature of the skill as documented in the README.
Logic is implemented in src/plugin.ts and uses a local state file .opencode/state/continual-learning.json for persistence.
[SAFE]: File System Management
The skill writes a local copy of its definition to .opencode/skills/continual-learning/SKILL.md to ensure availability within the project context. This file access is limited to the project directory.
[SAFE]: Data Ingestion and Sanitization
The skill processes conversation history to update AGENTS.md (Category 8 surface).
Mitigation: The SKILL.md instructions include a strict 'Exclusions' policy, explicitly forbidding the storage of secrets, tokens, credentials, and private personal data.
The workflow requires the AI to merge and deduplicate entries, focusing on high-signal, reusable information only.

continual-learning