continuous-learning

To extract reusable patterns from agent sessions, the skill identifies recurring corrections, workflows, and gotchas, then codifies them as maintainable framework additions. This prevents repeated mistakes by turning session insights into classified, validated knowledge.

Codifying session interactions into framework additions involves classifying patterns such as constraints, workflows, and shortcuts, then guiding you through conflict checking, validation, and confidence scoring. This ensures observed knowledge is safely integrated into the agent framework.

Confidence scoring for promoting session patterns evaluates observed reliability before moving project-level patterns to a global framework. The workflow applies confidence thresholds to ensure only validated, high-certainty knowledge is promoted across the agent framework.

You should run continuous-learning when agent sessions contain user corrections, repeated workflows, or discovered gotchas indicating potential reusable knowledge. Capturing these session insights prevents recurring mistakes and builds a maintainable framework.

Classifying constraints and automations requires conflict checking to prevent contradictions within the agent framework. The skill guides you through addition creation and validation, ensuring new patterns align with existing framework rules before integration.

Session patterns that can be extracted and codified include constraints, workflows, shortcuts, personas, automations, and lessons. The skill classifies these recurring insights from agent sessions and manages their promotion based on confidence scoring.