SkillDistiller

Conversation pattern extraction pulls expert guidance from chats by identifying four behavioral types: corrections, questions/probes, quality gates, and analysis modes, converting them into a permanent replayable skill file.

Taxonomy-based extraction distinguishes conversation guidance by categorizing extracted patterns into corrections, questions/probes, quality gates, and analysis modes, ensuring the resulting skill teaches behavioral dispositions rather than rigid rules.

Yes, SkillDistiller supports cross-conversation pattern confirmation by analyzing stored transcripts from a session id or file path, allowing you to validate expert behavioral guidance across multiple sessions before generating a skill.

Verify generated skills against session evidence using the built-in verification support, which checks the SKILL.md behavioral dispositions against the original conversation transcript to ensure accurate pattern extraction.

Yes, collaborative co-editing is supported during skill distillation; the tool presents extracted conversation patterns one at a time so you can approve, edit wording, adjust generalization level, or reject them before final skill generation.

To run transcript analysis for skill distillation, you need the qmd dependency installed in your environment, along with access to Claude Code conversation transcripts from a current session, session id, or file path.