rekol-initlisted

# rekol-init The assistant-led onboarding interview — the **primary** way a fresh REKOL install gets set up. **You** (the user's own Claude) run it: a short, adaptive five-step conversation that turns an empty install into searchable recall and, optionally, curated understanding. You shell out to mechanical `rekol` commands for the I/O and do the bootstrap reasoning yourself, all in this one context. The mechanical `rekol init` CLI is the scriptable fallback; this skill is the real onboarding. Run it when the user says `/rekol init`, "set up my rekol memory", "rekol initialize my memory", or otherwise asks to get started. ## The two honest layers (the spine of the whole interview) Frame everything around these two — they are what makes onboarding feel honest rather than magical: - **Recall** — your Claude Code history and knowledge files, made *searchable now*. Indexing is mechanical and immediate. This is step 1 (sessions) and step 2 (knowledge dirs). - **Understanding** — recurring instructions distilled into *curated, ambient memory* that grows over time. This is the bootstrap (step 3), and it keeps growing as you work (the capture protocol in the `rekol` skill). ## Non-negotiables (read before you start) - **Nothing is indexed or captured without an explicit yes.** Every step is opt-in. You may run the *detection* commands freely (they only read), but you run an *ingesting* command (`session-index`, `include add`, `bootstrap`, `capture`) only after th