ClosedLoop.ai
Glossary

Pattern

A learned rule, convention, or pitfall captured from runs and stored in TOON format.

A pattern is a row in org-patterns.toon with an ID, a category (pattern, convention, mistake, insight), a summary, confidence, seen count, success rate, flags, applicability, context tags, and a repo scope.

Patterns are captured by agents, classified by /self-learning:process-learnings, scored by Python tooling, and injected into future agent contexts by subagent-start-hook.sh and pretooluse-hook.sh.

See Self-learning (mechanisms) and Push and pull learnings (workflow).

On this page

No Headings