Show HN: Skill-extractor turns coding agent transcripts into reusable skills (github.com)

A new tool called Skill Extractor has been announced, designed to mine reusable skills from the transcripts generated by coding agents, including Claude Code and OpenAI Codex. This tool scores each identified skill based on its confidence and utility, evaluating the success of the tasks from which the skills were derived. Only the highest-quality skills are presented for user review, allowing developers to approve, edit, or reject them through a VS Code panel or command line interface. Approved skills are saved as markdown files, enhancing collaboration and efficiency for both individual developers and teams. The significance of Skill Extractor lies in its ability to transform repetitive coding habits into codified skills, thereby streamlining the development process. Crucially, it ensures privacy by keeping all operations local on the user’s machine, with built-in mechanisms for redacting sensitive information before sending any data to external models. Furthermore, the tool adopts a learning loop that adjusts future mining efforts based on user feedback, continually improving its effectiveness over time. This approach not only fosters a shared repository of skills but also facilitates team governance and security in developing AI-powered coding solutions.