How to use Claude to recover blocked Kindle highlights
A Claude Code skill recovered export-blocked Kindle highlights using local position metadata, Cloud Reader images, and Apple Vision OCR, producing cited Markdown notes.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Using [Mac Kindle annotation database] and my own logged-in [Chrome Cloud Reader session], recover [book] highlights for personal notes. Read highlight positions, capture relevant page images, OCR locally with Apple Vision, match known prefixes and lengths, and write location-cited text to [Markdown output].
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- l3a0
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports 2,432 highlights across four books, including 815 export-blocked highlights, all recovered verbatim with location citations into one Markdown file per book.
- Human involvement
- The post does not describe manual review steps; the author reports recovery results and position residuals.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Read character-precise highlight positions from the Mac Kindle SQLite database.
- Capture Cloud Reader page images through canvas.
- Run local OCR on the captured images with Apple Vision.
The author built a Claude Code skill after an Amazon notebook export omitted or truncated many of their Kindle highlights. In one book, 283 passages were cut off mid-sentence and 180 appeared only as location numbers. The post describes recovering these personal annotations through a combination of local metadata and rendered pages.
The Mac Kindle app’s SQLite database supplied character-precise start and end positions for each highlight. The skill captured Cloud Reader pages as images through canvas and used Apple Vision for local OCR. With the known positions, recovery became a matching and cutting operation: find the known prefix, then cut the text to the known length.
Recovered passages went into one Markdown file per book with location citations. Across four books, the author reports 2,432 highlights, including 815 blocked by export limits, all recovered verbatim. They also report position residuals of zero to two characters, with a median of zero to one.
The setup requires macOS, the Mac Kindle app’s data files, AppleScript browser control, Apple Vision, the user’s logged-in Chrome, and Claude Code. The author says it reads only their own annotations and does not touch DRM.
Who it's for: Mac users organizing their own Kindle annotations into location-cited personal notes.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · l3a0
Frequently asked questions
How did the skill determine highlight boundaries?
The Mac Kindle app’s SQLite database supplied exact start and end positions. Recovery used a known prefix and the known length to cut OCR text.
What platform and account access were required?
The post specifies macOS, AppleScript browser control, Apple Vision OCR, Mac Kindle app data files, the user’s own logged-in Chrome, and Claude Code.
More details
Remaining steps
- Find known prefixes and cut the text to the known highlight lengths.
- Write recovered highlights with location citations into Markdown files.
About this account
Claude Code drove a skill combining local highlight metadata, browser page capture, OCR, and position-based text recovery.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=49424759. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: l3a0 on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Mac Kindle app · SQLite · Kindle Cloud Reader · Apple Vision · Chrome · AppleScript
Kindle highlights · OCR · Personal notes
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.