https://github.com/albertnahas/audible-second-brain
Personal Audible library second-brain. Triage your library + wishlist with an evidence-based scorer, an interactive dashboard, and a heuristic rubric that calibrates to your own completion patterns over time. A Claude Code plugin.
https://github.com/albertnahas/audible-second-brain
audible audiobooks claude-code claude-code-plugin knowledge-management personal-productivity second-brain triage
Last synced: 13 days ago
JSON representation
Personal Audible library second-brain. Triage your library + wishlist with an evidence-based scorer, an interactive dashboard, and a heuristic rubric that calibrates to your own completion patterns over time. A Claude Code plugin.
- Host: GitHub
- URL: https://github.com/albertnahas/audible-second-brain
- Owner: albertnahas
- License: mit
- Created: 2026-04-27T17:05:14.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2026-04-27T18:26:23.000Z (about 2 months ago)
- Last Synced: 2026-04-27T19:10:50.379Z (about 2 months ago)
- Topics: audible, audiobooks, claude-code, claude-code-plugin, knowledge-management, personal-productivity, second-brain, triage
- Language: JavaScript
- Homepage: https://github.com/albertnahas/audible-second-brain
- Size: 75.2 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
Awesome Lists containing this project
README
# Audible Second Brain
A Claude Code plugin that turns your Audible library into a working second brain: an evidence-based scorer, an interactive triage dashboard, and a heuristic rubric that calibrates to *your own* completion patterns over time.
This is **promptware** — most of the value is in the prompts (skills) and the heuristic rubric. The plugin scaffolds a tiny Python scorer and a vanilla-JS dashboard into your workspace, then orchestrates everything via natural language inside Claude Code.
## What it does
| Skill | What it does | When you'd use it |
|---|---|---|
| `bootstrap` | Installs `audible-cli`, walks OTP login, scaffolds scorer + dashboard from generic templates | First time, or in a fresh workspace |
| `sync` | Re-exports library + wishlist, classifies new books, regenerates scored data, surfaces deltas | After buying / finishing books, or weekly |
| `classify` | Routes each book into a topic cluster using Claude Haiku via the headless CLI; caches results by ASIN | After bootstrap, after editing the cluster taxonomy, or to fix miscategorizations |
| `recommend` | Proposes new candidates aligned with your rubric | "What should I read next?" |
| `calibrate` | Re-derives HIGH_TRUST authors, anti-patterns, length cliffs, cluster weights from your own completion data | Once you have ≥ 20 finished books, or after a year of new data |
| `triage` | Walks you through PASS / LATER / KEEP decisions, persists to dashboard + audit log | "Help me clean up my list" |
A SessionStart hook surfaces a one-line freshness reminder when your snapshot is > 14 days old or when you've crossed the calibration threshold.
## Install
```bash
# In Claude Code:
/plugin marketplace add albertnahas/audible-second-brain
/plugin install audible-second-brain@albertnahas/audible-second-brain
```
Then in a fresh directory:
```
> /audible-second-brain:bootstrap
```
The skill takes it from there.
## How the rubric works
Every owned and wished book is scored on four axes:
1. **Author trust** — books from authors you reliably finish get a positive score; authors whose long-form works you've never opened get a negative score.
2. **Cluster fit** — books are routed to a topic cluster by Claude Haiku (via the headless `claude` CLI), based on title + subtitle + author + Audible genres read together. Each cluster has a weight reflecting how well that topic lands for you. Falls back to a conservative regex if no LLM classification exists yet.
3. **Length fit** — your personal "abandon cliff" (the runtime above which completion rates collapse) sets a soft penalty on long books unless they're in a high-affinity cluster.
4. **Anti-patterns** — formats or publishers where you have a track record of buying-but-not-finishing get a penalty (e.g., publisher series you've collected but never opened).
Each book then gets a category:
- Library items → **PASS** / **LATER** / **KEEP**
- Wishlist items → **CUT** / **MAYBE** / **KEEP**
The starter rubric ships with generic clusters that cover common non-fiction territory and empty author/anti-pattern lists. The **`calibrate`** skill replaces those defaults with values derived from your own data:
- Trusted authors → those with ≥ 2 finished books and a completion ratio ≥ 0.6
- Anti-patterns → formats or publishers where you own ≥ 3 books and finished 0
- Cluster weights → re-balanced so the topics you actually finish get bonus, the ones you don't get neutral
- Length cliff → derived from your per-runtime-band completion ratios
This means the scorer becomes more accurate over time without you having to think about it.
## Repository layout
```
audible-second-brain/
├── .claude-plugin/
│ └── plugin.json
├── skills/
│ ├── bootstrap/SKILL.md
│ ├── sync/SKILL.md
│ ├── classify/SKILL.md
│ ├── recommend/SKILL.md
│ ├── calibrate/SKILL.md
│ └── triage/SKILL.md
├── hooks/
│ └── hooks.json
├── scripts/
│ ├── check-freshness.sh
│ └── check-freshness.py
├── bin/
│ └── audible-sync-cli # cron-friendly headless sync
├── templates/
│ ├── _score.starter.py
│ ├── _classify.starter.py
│ ├── preferences.starter.md
│ ├── CLAUDE.template.md
│ ├── .gitignore.template
│ ├── serve.sh
│ ├── dashboard/
│ │ ├── dashboard.html
│ │ └── dashboard.js
│ └── scripts/
│ └── regenerate-md.py
├── docs/
│ ├── architecture.md
│ └── design-decisions.md
├── README.md · LICENSE · CHANGELOG.md · CONTRIBUTING.md · .gitignore
```
## Privacy
- Audible auth is stored only in `~/.audible/` (managed by `audible-cli`, never read by this plugin).
- Library + wishlist data stays in your workspace as plain JSON / CSV / MD.
- The plugin makes no outbound network calls beyond what `audible-cli` does to refresh exports.
## Requirements
- Python 3.9+
- [`audible-cli`](https://github.com/mkb79/audible-cli) ≥ 0.3 (the bootstrap skill installs it via `pipx`)
- Claude Code 2.x or newer (the `classify` skill uses the bundled `claude` CLI in headless mode)
## Headless / cron usage
```bash
# Refresh once a week, Monday 9am
0 9 * * 1 /path/to/your/workspace && /path/to/plugin/bin/audible-sync-cli
```
The CLI wrapper handles re-export, re-score, regenerate-md, and snapshot bookkeeping without needing a Claude session.
## Contributing
See [`CONTRIBUTING.md`](CONTRIBUTING.md). Issues and PRs welcome — especially calibration heuristics and dashboard refinements.
## License
MIT — see [`LICENSE`](LICENSE).