CLI-first image verification skill for Skillshare.

content-checker checks one or more images against text requirements using Gemini Flash and returns machine-readable JSON for iterative regenerate/review loops.

• Verifies single or batch images
• Accepts requirements inline or from file
• Returns structured pass/fail output with confidence and gap analysis
• Generates concrete adjustment guidance and a reprompt candidate
• Threshold + hard-fail logic for automation

• Python 3.10+
• uv
• GEMINI_API_KEY environment variable

export GEMINI_API_KEY="your_api_key"

uv run scripts/check_images.py \
  -i /path/to/image.png \
  --requirements-text "Logo is visible in top-left and CTA is readable" \
  --json

Use one of these (or combine):

1. --requirements-text "..."
2. --requirement "id::text" (repeatable)
3. --requirements-file requirements.json

JSON format supports string or object entries:

[
  "Hero image includes product",
  {"id":"cta","text":"CTA text is readable","weight":1.0,"must":true}
]

uv run scripts/check_images.py \
  -i examples/sample.jpg \
  --requirements-file examples/requirements.json \
  --threshold 0.85 \
  --json \
  --out /tmp/check-result.json

• 0: evaluation ran and passed
• 3: evaluation ran but failed threshold/rules
• 1: model/runtime error
• 2: input/validation error

Install from this repo:

skillshare install github.com/nerveband/content-checker --track
skillshare sync

Or use local source:

skillshare install /path/to/content-checker-skill
skillshare sync

• SKILL.md: skill trigger and usage guidance
• scripts/check_images.py: CLI checker implementation
• references/output-schema.md: output contract