setuplisted

# /mneme:setup — One-time onboarding checkup for a repository The ENTRY for a new user or a new project. One run per repository: diagnose the ALREADY WIRED mneme surface with zero questions, render one ternary report, offer ONLY the applicable fixes — each behind its own digit — then walk the FIRST note through the human staging gate so the remember → human gate → recall loop is demonstrated on something real, and end with a short "what's next" map. It is a checkup, not a wizard: a fully green project gets a short green report and ZERO questions. FTS-first honesty is the skill's stance: recall without an embedder runs in degraded mode over FTS — that is a DECLARED working entry level ("works immediately; vector is an upgrade"), never an error to apologize for. The upgrade fork (Ollama, or any OpenAI-compatible endpoint) is offered, not pushed. Diagnosis reads the ENGINE's own signals through the wired surface (`stats`, `recall`, git, the config file) — it does NOT re-probe the embedder and does NOT duplicate the engine's doctor internals. When the engine later exposes a doctor MCP tool, ONLY the DIAGNOSE section below changes. ## Arguments `/mneme:setup` — no arguments; runs against the current repository. ## Permissions (VIOLATION = ABORT) - Read / Grep / Bash: YES — for DIAGNOSE probes only (`git rev-parse`, reading `.mneme.json`, grepping the project's `CLAUDE.md`). Bash MUTATES nothing except the digit-consented first commit below. - Write / Edit into the proje