self-improvement

Agent errors and learning opportunities are logged by appending structured markdown entries to a `.learnings/LEARNINGS.md` file, capturing command failures, user corrections, and knowledge gaps for future reference.

AI agent feedback loops are captured through automated hook integration that detects command failures and unexpected behaviors, logging their context and suggested fixes to facilitate continuous workflow improvement.

Discovered best practices are promoted by elevating logged insights to permanent project memory files such as `CLAUDE.md`, `AGENTS.md`, or `TOOLS.md`, ensuring recurring corrections persist across future agent sessions.

Recurring error fixes can be extracted as new reusable skills, transforming repeated command failures and documented knowledge gaps into structured capabilities that prevent identical errors in future development workflows.

This continuous improvement approach works without external dependencies, relying solely on internal scripts, references, and assets to manage structured logging of feature requests, errors, and best practices.

Structured learning logs should not be used for ephemeral errors or one-off command failures that lack reusable context, as the overhead of logging outweighs the benefit when insights cannot be promoted to permanent memory.