claude-history-ingest

The skill processes both CLI session data and desktop local-agent-mode sessions, extracting user and assistant content to write structured wiki pages with provenance and confidence metadata.

Yes, it supports append modes using delta detection via a vault manifest. This allows updating only new or modified items from conversation transcripts without reprocessing the entire history.

Conversation mining for knowledge base distillation is the process of extracting reusable knowledge from past chats. It skips secrets and verbatim logs, grouping extracted insights by topic into concepts, projects, and entities.

Yes, you can extract workflow patterns from Claude audit logs. The ingestion process optionally mines these audit logs to distill workflow patterns directly into structured wiki pages.

The best way to organize past Claude chats into a searchable knowledge base is by clustering extracted topics into mapped wiki pages. This includes projects, concepts, skills, and entities while keeping provenance and confidence signals.

Wiki ingestion skips verbatim chat logs during data distillation to maintain a privacy-first approach. It focuses on extracting and synthesizing reusable knowledge rather than storing sensitive raw conversation transcripts.