What problem does it solve? Setting up a persistent memory and semantic search backend (gbrain) for a coding agent involves many manual steps: installing the CLI, initializing a local PGLite or Supabase store, registering the MCP server, and defining trust policies per git remote. This Skill automates that entire flow with a single command and is idempotent, so re-running it repairs missing pieces instead of duplicating work. ## Core Features & Use Cases - End-to-end gbrain setup: Installs the gbrain CLI, initializes a local PGLite or Supabase brain, and registers the MCP server (local stdio or remote HTTP with bearer token verification). - Transcript and memory ingest: Ingests Claude Code and Codex session transcripts, learnings, timelines, and design docs into gbrain, with optional gitleaks secret scanning and per-repo deny policies. - Trust policy and artifacts sync: Captures per-remote trust policy, provisions a private artifacts repo for cross-machine sync, and writes a GBrain configuration block to CLAUDE.md. - Use Case: A developer on a new machine runs the setup once and the agent gains semantic code search (gbrain search, code-def, code-refs) plus recall of prior sessions, with a GREEN/YELLOW/RED verdict block confirming each component. ## Quick Start Ask the agent to set up gbrain on this machine, including installing the CLI, initializing the brain store, and registering the MCP server.
