prompt-intelligence

To load learned behaviors at agent session start, invoke the /prompt-intelligence command. It extracts accumulated user corrections from cortex annotations or a knowledge.jsonl fallback to build a session-wide behavior policy block.

Yes, you can sync learned agent behaviors to Obsidian. The skill automatically translates accumulated user corrections into learned patterns and syncs them to Obsidian, providing cross-agent visibility and review.

A learned behaviors context block is a translated compilation of past user corrections that shapes AI agent actions. You need it when you want an agent to automatically apply previously learned rules across a new session.

You do not strictly need a cortex_lessons feed to apply past user corrections. The skill requires either a cortex_lessons feed or a knowledge.jsonl file as a fallback to load and categorize the accumulated behavioral data.

You can categorize and bootstrap learned behaviors by running the skill's dedicated phases. It processes the cortex data through loading, categorizing, building, and bootstrapping phases to establish the rules for the current session.

The knowledge.jsonl fallback provides static session management by reading past corrections locally, but lacks the dynamic, real-time annotation stream provided by a live cortex_lessons feed for continuous behavior loading.