The Agent Skills Directory

[DATA_EXFILTRATION]: The skill logic directs the agent to read user-provided credentials from config/secrets.md and transmit them to an unverified third-party endpoint (https://api.apicore.ai/v1/images/generations) rather than a direct, well-known AI provider.
[DATA_EXFILTRATION]: The documentation in tips/image-upload.md provides a command template for uploading files to catbox.moe using curl -F "fileToUpload=@path". This pattern is highly susceptible to exfiltrating sensitive local files (e.g., SSH keys or environment files) if the path parameter is manipulated via prompt injection.
[EXTERNAL_DOWNLOADS]: The skill interacts with several external, non-whitelisted domains including api.apicore.ai, catbox.moe, and litterbox.catbox.moe for its primary operations.
[COMMAND_EXECUTION]: The skill relies on constructing and executing curl commands to communicate with remote APIs and upload services.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it processes untrusted external data (image URLs and user prompts) that are interpolated directly into shell commands.
Ingestion points: Image URLs and prompt text processed in SKILL.md.
Boundary markers: Absent; no delimiters are used to separate user input from the command structure.
Capability inventory: Access to curl with the ability to read local files via the @ prefix.
Sanitization: Absent; no validation or escaping is performed on the user-provided URLs or descriptions.

gemini-image