What problem does it solve? When you ask "how did we solve this last time" or "what was the decision on X", the answer often lives in old AI conversation logs rather than in any committed file. This Skill teaches the AI how to query the trellis mem CLI to search, slice, and extract past Claude Code, Codex, Grok, Pi, and ZCode session dialogue so prior decisions and solutions can be recovered instead of re-litigated. ## Core Features & Use Cases - Keyword search across sessions: Find past sessions mentioning a symptom, decision, or topic, scoped to the current project or globally across the machine. - Phase-based extraction: Slice a session by Trellis task boundaries (--phase brainstorm / --phase implement) to recover planning discussions or execution loops. - Context drilling: Pull the top-N hit turns plus surrounding context from a session with a token budget, or dump cleaned dialogue for review. - Use Case: You hit a timeout bug that feels familiar. The AI runs trellis mem search "timeout" --global, finds a session from last month, extracts the relevant turns, and quotes the prior fix with the session id so you can verify it. ## Quick Start Ask the AI to search past conversation history for how a similar problem was solved before, for example by saying "check what we decided about the database choice last time".
