audible-second-brain

First-time setup of an Audible second-brain workspace. Use when the user wants to "set up audible-second-brain", "install audible cli", "bootstrap my book library", "start tracking my Audible books", or is in a fresh directory and asks Claude to begin organizing their Audible library. Walks through audible-cli installation, OTP login, the first library + wishlist export, and scaffolds preferences.md / _score.py / dashboard from generic starter templates.

Classify books into topic clusters using Claude Haiku via the headless CLI. Use when the user says "classify my library", "fix the clusters", "categorize my books", "re-classify everything", or after `bootstrap` / `sync` notices new uncached books. Writes a per-ASIN cache (`classifications.json`) so repeated runs are cheap and idempotent.

Refresh Audible library and wishlist exports, regenerate scored JSON, and update the dashboard. Use when the user says "sync my library", "refresh books", "update audible data", "I bought a new book", "I finished a book", "regenerate dashboard", or after a SessionStart hook reports the snapshot is stale (>14 days old).

Re-derive the user's personal scoring rubric (HIGH_TRUST authors, ANTI_PATTERNS, CLUSTER_RULES, length preferences) from observed completion data. Use when the user says "calibrate my rubric", "personalize the scorer", "learn my taste", "update my preferences from data", or after enough new completion signal accumulates (≥ 20 finished books since last calibration, or first time after bootstrap).

Propose new audiobook candidates aligned with the user's preferences.md rubric. Use when the user says "what should I read next", "find me books like X", "recommend audiobooks", "suggest something on topic Y", "fill my wishlist with quality candidates". Returns scored candidates with rationale, ready to add to the wishlist.

Guide the user through interactive PASS/LATER/KEEP review of library and wishlist items. Use when the user says "triage my wishlist", "clean up my library", "decide what to keep", "what should I cut", "let's go through my list". Surfaces highest-leverage decisions first (long unfinished commitments, low-score wishlist items) and persists decisions to dashboard localStorage and an audit log.