What problem does it solve? Locating a past DSH conversation by its title or something said inside it normally requires manually traversing session directories and decompressing multi-frame zstd files one by one. This Skill replaces that with a single script that searches the session index, scans message bodies, and reports content and token usage. ## Core Features & Use Cases - Title Search: Query the session_projcache.json index by title keyword, with filters for project path, date range, and sorting by time, tokens, or turns. - Full-Text Body Search: Grep inside session message content across the most recent 30 sessions, or all sessions with the --full flag. - Content and Stats Reading: Read a session's messages by ID or view duration, turns, tool calls, and complete token usage (uncached input, cache read, output, cache write). - Use Case: You remember discussing "context overflow" in an earlier conversation but not its title. Run a body search to find the matching sessions with message snippets, then open the full content or check its token consumption by session ID. ## Quick Start Ask the agent to find the previous DSH conversation about a topic you remember and show how many tokens that session consumed.
