self-improving-meta

To prevent prompt drift in agent infrastructure, this skill logs meta-learnings and applies targeted fixes to core prompt files. It detects drift across sessions and promotes widely applicable improvements into shared infrastructure for consistent agent behavior.

Resolving rule conflicts and context bloat involves centralizing meta-learnings into stable files and hardening core prompt files. This process detects infrastructure degradation and propagates vetted fixes across shared components to maintain optimal agent performance.

You can propagate infrastructure fixes safely by using hook validation and meta-extraction workflows. This ensures that vetted improvements to core prompt files, memory policies, and hooks are inherited consistently across all skills without breaking existing functionality.

Yes, you can fix memory degradation and misinterpretations by operating across all sessions to detect these issues. It applies improvements to memory policies and core infrastructure files, ensuring agents handle context and instructions accurately over time.

The skill improves existing core prompt files such as AGENTS.md, CLAUDE.md, SOUL.md, and TOOLS.md. It consolidates feature requests and learnings into files like LEARNINGS.md, hardening the current agent setup rather than generating an infrastructure from scratch.

You should run a meta-learnings review after major changes to capture infrastructure improvements and propagate fixes. This workflow detects prompt drift, rule conflicts, and memory degradation, ensuring your shared files remain stable across all active sessions.