seemenot PII Redaction Skill

Redact personally identifiable information (PII) from images using local OCR. No cloud APIs—everything runs locally using Tesseract.

Prerequisites

Tesseract OCR must be installed:

• macOS: brew install tesseract
• Linux (Debian/Ubuntu): sudo apt install tesseract-ocr
• Windows: winget install UB-Mannheim.TesseractOCR

Install the CLI from GitHub:

uv tool install git+https://github.com/waldekmastykarz/seemenot

Or run without installing:

uvx --from git+https://github.com/waldekmastykarz/seemenot seemenot <image> -r <text>

Core Workflow

Basic Redaction

Redact specific text strings from an image:

seemenot <image> -r <text-to-redact> [-r <more-text>...]

Output is saved as <image>-nopii.<ext> by default.

Common Options

Option	Short	Description
--output	-o	Custom output path
--redact	-r	Text to redact (repeatable)
--exclude	-e	Text to exclude from detection (repeatable)
--redact-style		box (default), blur, or pixelate
--types		Auto-detect PII types: email, ip, jwt
--interactive	-i	Confirm each detection before redacting
--dry-run		Preview detections without modifying
--json		Output as JSON (with --dry-run)
--from-manifest		Apply redactions from JSON file
--verbose	-v	Show detection details

Typical Tasks

Redact Specific Text

When the user specifies exact strings to redact:

seemenot screenshot.png -r "waldek" -r "mypassword" -r "secret-hostname"

Auto-Detect PII Patterns

Automatically find and redact emails, IPs, and JWTs:

# Detect all supported types
seemenot screenshot.png

# Only specific types
seemenot screenshot.png --types email,ip

# Combine with explicit redactions
seemenot screenshot.png -r waldek --types jwt

Exclude False Positives

Prevent specific text from being redacted:

seemenot screenshot.png -r waldek -e "waldek-public-repo"

Choose Redaction Style

# Black box (default)
seemenot screenshot.png -r secret --redact-style box

# Gaussian blur
seemenot screenshot.png -r secret --redact-style blur

# Pixelate
seemenot screenshot.png -r secret --redact-style pixelate

Preview Before Redacting

Dry run shows what would be detected without modifying:

seemenot screenshot.png -r waldek --dry-run -v

Interactive Review

Review each detection before redacting:

seemenot screenshot.png -r waldek -i

Manifest Workflow (Fine-Grained Control)

For complex images requiring manual review:

# 1. Generate detection manifest
seemenot screenshot.png -r waldek --dry-run --json > detections.json

# 2. Edit detections.json to remove false positives

# 3. Apply from manifest
seemenot screenshot.png --from-manifest detections.json

Execution Guidelines

1. Always use absolute paths for images to avoid path resolution issues
2. Preview with --dry-run when uncertain about what will be detected
3. Use -v (verbose) to show detection details for verification
4. Combine -r and --types when both explicit strings and pattern detection needed
5. Use --exclude for known false positives rather than editing manifests
6. Prefer --redact-style blur when context around redacted content should be partially visible
7. Prefer --redact-style pixelate for a less aggressive anonymization

Supported PII Types

Type	Examples	Notes
email	john@example.com	Standard email pattern
ip	192.168.1.1, 10.0.0.1:8080	Excludes localhost 127.x.x.x
jwt	eyJhbGciOiJ...	JSON Web Tokens

For usernames, hostnames, API keys, or other text—use -r <text>.

Error Handling

• Exit 0: Success or no PII detected
• Exit 1: Error during detection or redaction
• Exit 2: Invalid input (e.g., unknown PII type)

Check stderr for error messages. Use -v to debug detection issues.

Example Scenarios

Clean a screenshot for documentation

seemenot docs/screenshot.png -r "john.doe@company.com" -r "192.168.1.100" -o docs/screenshot-clean.png

Anonymize terminal output

seemenot terminal.png --types email,ip -r "$(whoami)" --redact-style blur

Batch processing with manifest

# Generate, review, apply
seemenot image.png -r sensitive --dry-run --json > manifest.json
# (manually edit manifest.json)
seemenot image.png --from-manifest manifest.json -o image-safe.png