harness-improvement

You can integrate session learnings by applying the harness-improvement skill, which captures recurring outcomes and explicit requests to update your AI harness. This ensures continuous improvement by modifying guidance files and agent structures based on what was learned during previous AI interactions.

Yes, you can update skills, agents, and commands across Claude, Codex, Cursor, and OpenCode platforms. The harness-improvement skill performs cross-platform adaptation, ensuring your AI workflow updates and guidance files are applied consistently across all four supported environments.

The best way to record a new learning is to run the harness-improvement skill, which captures session outcomes and explicit requests to update your harness structure. This updates your skills and guidance files automatically, ensuring continuous improvement of your AI workflows without manual file editing.

Yes, you need knowledge of harness structure and user requests to effectively use the harness-improvement skill. Understanding how skills, agents, and guidance files are organized ensures that explicit requests and learnings are correctly integrated into your AI workflow for successful continuous improvement and cross-platform adaptation.

Cross-platform harness adaptation ensures AI workflow updates and skill modifications are consistently applied across Claude, Codex, Cursor, and OpenCode. Without it, learnings and explicit requests would need manual translation for each platform, breaking the continuous improvement cycle and fragmenting agent management.