The Agent Skills Directory

[SAFE]: The skill's primary purpose is image localization, and all observed behaviors in the instructions and the Python script are consistent with this stated goal. No obfuscation or malicious logic was found.
[EXTERNAL_DOWNLOADS]: The skill uses the urllib library to download images from URLs specified within Markdown documents. The implementation includes safety features such as maximum file size limits and an optional domain whitelisting mechanism (--allowed-host) to restrict network access to trusted sources.
[COMMAND_EXECUTION]: The provided script manages local files and directories to store the localized packages. It includes robust sanitization logic to strip invalid characters and prevent path traversal attacks by ensuring output paths are safely derived from input filenames.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Markdown data which could contain malicious instructions. The analysis confirms the following safety posture:
Ingestion points: The localize_markdown function in illustrations_local_storage.py reads the full text of external Markdown files.
Boundary markers: The SKILL.md file contains explicit instructions to treat the input as data and disregard any embedded commands.
Capability inventory: The script has network access for image fetching and file system access for writing the output package.
Sanitization: The script uses regular expressions to isolate URLs, ignores content within code fences, validates image integrity via magic bytes before processing, and normalizes all images to JPEG using the Pillow library.

illustrations-local-storage