The Agent Skills Directory

[DATA_EXPOSURE_&_EXFILTRATION]: The skill accesses authentication tokens and API URLs from the user's home directory (~/.rebyte.ai/auth.json). This is a standard and safe practice for CLI tools managed by a specific vendor (rebyteai) to maintain persistent authentication.
[EXTERNAL_DOWNLOADS]: The skill performs network operations via urllib.request to a dynamic relay_url fetched from the vendor's local configuration. Since these requests are directed to the vendor's infrastructure for the primary purpose of the skill (image generation), this is considered safe vendor functionality.
[COMMAND_EXECUTION]: The SKILL.md file provides clear instructions for the AI agent to execute the included Python script. All parameters are passed as arguments to a controlled CLI script rather than raw shell interpolation of user input, minimizing injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for external data via the --input flag which reads local image files. However, the data is base64 encoded and sent to an API, and the skill does not interpolate this untrusted content back into prompt instructions, limiting the risk of indirect injection.

image-generation