The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interfaces with system utilities by calling gitleaks (in scripts/secret_scan.py) and tesseract (in scripts/image_redact.py and scripts/selective_ocr.py) using subprocess.run. These calls utilize a list of arguments to avoid shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: Maintenance scripts such as scripts/update_secret_rules.py fetch configuration data and license information from the official Gitleaks repository on GitHub. This is a standard procedure for keeping security rules up to date.
[EXTERNAL_DOWNLOADS]: The scripts/replicate_finish.py script interacts with the Replicate API for optional generative image editing. This feature is consent-gated, requiring the user to provide explicit approval flags and an API token via environment variables before any network communication occurs.
[INDIRECT_PROMPT_INJECTION]: Because the skill is designed to process external text and images, it possesses an attack surface for indirect prompt injection. Ingestion points include scripts/text_redact.py and scripts/image_redact.py. The skill documentation includes specific instructions for the agent to disregard any commands embedded within the processed content. Capability inventory includes subprocess calls for scanning and OCR, and network operations in replicate_finish.py. The skill implements redaction logic in text_redact.py as a primary function, which acts as a sanitization layer.

anonymizer