Skip to content
Skills Library
AI & agents · Analysis

squint-reader

/squint-reader

Reveal hidden text in blur-illusion or hybrid images that vision models misread. Use when an image looks like blurry, glowing, or overlapping outlines, when a user says an AI "cannot read" an image, or when the visible text seems wrong or nonsensical. Low-pass filtering (downscale then upscale) exposes the true low-frequency message.

Licence
MIT
Last repository change
21.07.2026

Assessment and security

95/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content30 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 21.07.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Reveal hidden text in blur-illusion or hybrid images that vision models misread. Use when an image looks like blurry, glowing, or overlapping outlines, when a user says an AI "cannot read" an image, or when the visible text seems wrong or nonsensical. Low-pass filtering (downscale then upscale) exposes the true low-frequency message.

Triggers from the description

cannot read
Original author description

Reveal hidden text in blur-illusion or hybrid images that vision models misread. Use when an image looks like blurry, glowing, or overlapping outlines, when a user says an AI "cannot read" an image, or when the visible text seems wrong or nonsensical. Low-pass filtering (downscale then upscale) exposes the true low-frequency message.

How it works

  1. When to use
  2. Method
  3. Reporting
Original SKILL.md headings · 40 lines, 240 words
SKILL.md preview
---
name: squint-reader
description: Reveal hidden text in blur-illusion or hybrid images that vision models misread. Use when an image looks like blurry, glowing, or overlapping outlines, when a user says an AI "cannot read" an image, or when the visible text seems wrong or nonsensical. Low-pass filtering (downscale then upscale) exposes the true low-frequency message.
---

# SquintReader

## When to use

Reach for this when an image shows soft, neon, or concentric-outline letters, when the readable text looks like gibberish, or when the user claims a message is hidden that AI keeps reading wrong. These are hybrid images: the sharp high-frequency layer is a decoy, and the real message lives in the low frequencies that appear only when you squint or step back.

## Method

Apply a low-pass filter — downscale the image with area averaging, then upscale it back and raise contrast. The fine outlines average away and the bold hidden shapes remain.

### Preferred: use the package

```bash
squintreader INPUT.png                 # writes INPUT_revealed.png as a montage of squint strengths
squintreader INPUT.png -s 0.12 --ocr   # single reveal plus recovered text
```

Read the revealed image. Report the text that appears in the low-frequency layer, not the sharp original.

### Fallback: a few lines of Pillow

```python
from PIL import Image, ImageEnhance, ImageOps

image = Image.open("INPUT.png").convert("RGB")
width, height = image.size
for scale in (0.20, 0.12, 0.08):
    small = image.resize((max(16, round(width * scale)), max(9, round(height * scale))), Image.BOX)
    revealed = small.resize((width, height), Image.BICUBIC)
    revealed = ImageEnhance.Contrast(ImageOps.grayscale(revealed).convert("RGB")).enhance(1.6)
    revealed.save(f"revealed_{scale}.png")
```

Open each `revealed_*.png` and read the one where the hidden words are clearest. Smaller `scale` squints harder; add `ImageOps.invert` for light text on a dark background.

## Reporting

State the recovered message plainly and note it was the hidden low-frequency reading, distinct from the misleading sharp text.

Versions

The database has no other skills named squint-reader.

Database snapshot: 13.11.2026 · package version dd611386aa06 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100analyze-images
    Read, describe and answer questions about images (screenshots, renders, photos) using the vision MCP server backed by the three.ws vision pipeline. Use when the user wants OCR, a caption, alt text or a visual critique of an image URL or base64 image.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.