instincts

To capture recurring user preferences, you can record observed patterns and corrections during sessions into Forgetful memory, which tracks confidence and invocations to graduate proven instincts into reliable patterns automatically.

Graduating learned patterns into permanent skills requires meeting strict criteria: confidence must exceed 0.9, invocations must surpass 10, and the pattern must have no contradictions with existing rules or workflows.

You can review and export tracked session patterns by running the /instinct-status command to list them with confidence scores, then using the /instinct-export command to save the instincts to a JSON file.

Forgetful memory tracks successful approaches across workflows by storing observed corrections and patterns during active sessions, automatically reinforcing successful approaches while applying governance to avoid contradictions.

Yes, automatic recall and adaptation for recurring workflow corrections is enabled once an instinct proves reliable, allowing the system to automatically apply learned preferences across your active workflows.

The best way to evolve proven instincts into automated skills is using the /evolve command, which graduates eligible instincts through auditing and review once confidence and invocation thresholds are met.