recognize-captchalisted
Install: claude install-skill xsyetopz/dotclaude
<task>
Read the text in a CAPTCHA image with [ddddocr-rs](https://github.com/mzdk100/ddddocr-rs), a Rust implementation of ddddocr that recognizes CAPTCHA text offline with an ONNX model.
</task>
<approach>
The best CAPTCHA is one that never appears, so this OCR covers only the rare challenge that appears anyway. Do not make CAPTCHA solving your default strategy, because a site that keeps challenging you keeps escalating, and prevention avoids that. Work through these in order:
1. Use CloakBrowser to prevent challenges.
2. Use residential proxies for geographic legitimacy.
3. If a CAPTCHA still appears, try this offline OCR as a last resort.
4. For complex CAPTCHAs (image puzzles, reCAPTCHA v3), ask the user.
Offline OCR is the fallback of choice because it runs locally with no network latency, keeps images on the machine, has no per-solve cost, and works offline after the model download.
</approach>
<installation>
```bash
# Install the ddddocr-rs CLI
cargo install ddddocr-cli
# Download the ONNX model
mkdir -p ~/.local/share/ddddocr
curl -L -o ~/.local/share/ddddocr/ddddocr.onnx \
https://github.com/mzdk100/ddddocr-rs/raw/main/models/ddddocr.onnx
```
The script looks for the model in these locations, in order:
- `$DDDDOCR_MODEL_PATH`
- `~/.local/share/ddddocr/ddddocr.onnx`
- `~/.ddddocr/ddddocr.onnx`
- `/usr/local/share/ddddocr/ddddocr.onnx`
</installation>
<usage>
Recognize a CAPTCHA from a screenshot or image file; the script prints JSON:
```bash
# From a scree