one-skill-to-rule-them-all

Why we included this project

Task Observer runs inside Claude Code and similar agent setups, watching your work sessions and turning what it sees into improvements for your other skills. When you correct an AI's output or steer it a different way, it logs that as a signal that a skill could be clearer, and it also spots repeating patterns that deserve their own new skill. Because it observes every active skill, including itself, the methodology keeps refining itself as you use it. That matters for anyone who maintains a growing library of agent skills and wants them to stay aligned with how they actually work, rather than drifting stale. It's domain-agnostic and works with human-led sessions as well as autonomous agents, so it's a practical fit for small teams that want their skill collection to keep learning from daily use.