Skip to content
Back to search

case-narrator

Claude Code skill that turns business-school case PDFs into narrated MP3 read-throughs, using local Kokoro TTS

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/abhiroopvaddiparthi-a11y--case-narrator/badge.svg)](/listings/abhiroopvaddiparthi-a11y--case-narrator)

HTML

<a href="/listings/abhiroopvaddiparthi-a11y--case-narrator"><img src="/api/listings/abhiroopvaddiparthi-a11y--case-narrator/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

cp -r skill ~/.claude/skills/case-narrator

The first time the skill runs it sets itself up:

- **Mac:** `scripts/setup-mac.sh` uses only built-in tools (curl, shasum, tar), so there's no developer-tools popup. It downloads uv 0.13.6 (SHA-256 pinned), which installs a private Python 3.12 into `~/.case-narrator/python`, then runs `setup.py`. - **Windows:** `python scripts/setup.py` (Python 3.10+).

`setup.py` creates `~/.case-narrator/venv`, installs the exact versions in `requirements.txt` and downloads the voice model (about 120 MB), verifying its SHA-256. Everything lives in `~/.case-narrator`; deleting that folder plus `~/.claude/skills/case-narrator` uninstalls it.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 54 lines
  1. # Case Narrator
  2. Turns a business-school case PDF into a narrated MP3 for listening on the go. It is packaged as a
  3. Claude Code skill: drop the PDF into Claude, say "read this case to me", and the MP3 lands in
  4. iCloud Drive › Case Audio (Mac) or Music › Case Audio (Windows).
  5. ## How it works
  6. 1. `skill/scripts/extract_clean.py` pulls the text out of the PDF and strips mechanical noise
  7. (watermarks, running headers, page numbers, footnote markers). It keeps exhibits with their
  8. column layout.
  9. 2. Claude writes the narration following `skill/narration-guide.md` (light condensing, numbers
  10. spelled out for speech, a short gist for each exhibit).
  11. 3. `skill/scripts/synthesize.py` reads the narration with the local Kokoro voice model (default
  12. voice `bm_fable`) and encodes the MP3.
  13. Everything except step 2 runs on the computer. The PDF text goes to Claude, and nothing else
  14. leaves the machine.
  15. ## Folders
  16. | Path | What |
  17. |---|---|
  18. | `skill/` | The skill itself (source of truth) |
  19. | `test-fixtures/` | A fictional HBS-style case and the script that generates it |
  20. | `voice-samples/` | One-minute samples of six voices |
  21. | `work/` | Scratch output from test runs |
  22. | `models/`, `.venv/` | Local development copies; the installed skill uses `~/.case-narrator/` |
  23. ## Install / update the skill on this machine
  24. ```bash
  25. cp -r skill ~/.claude/skills/case-narrator
  26. ```
  27. The first time the skill runs it sets itself up:
  28. - **Mac:** `scripts/setup-mac.sh` uses only built-in tools (curl, shasum, tar), so there's no
  29. developer-tools popup. It downloads uv 0.13.6 (SHA-256 pinned), which installs a private
  30. Python 3.12 into `~/.case-narrator/python`, then runs `setup.py`.
  31. - **Windows:** `python scripts/setup.py` (Python 3.10+).
  32. `setup.py` creates `~/.case-narrator/venv`, installs the exact versions in `requirements.txt`
  33. and downloads the voice model (about 120 MB), verifying its SHA-256. Everything lives in
  34. `~/.case-narrator`; deleting that folder plus `~/.claude/skills/case-narrator` uninstalls it.
  35. ## Security notes
  36. - Downloads: uv and both model files are checked against pinned SHA-256 values; packages are
  37. pinned to exact versions (no known vulnerabilities per `pip-audit`, 2026-11-14).
  38. - The voice libraries contain no network code; only the PDF text goes to Claude.
  39. - The skill tells Claude to treat PDF text as content, never instructions, and to run only its
  40. own scripts.
  41. - Partial MP3s are built in `~/.case-narrator/tmp`, so a failed run never syncs junk to the phone.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.