The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/edit_artwork.py script uses subprocess.run() to execute commands defined in an external JSON configuration file.\n
The custom_edit function retrieves a command array from a user-specified adapter configuration and executes it on the system.\n
This mechanism allows for the execution of arbitrary binaries or scripts provided in the configuration.\n- [DYNAMIC_EXECUTION]: The skill implements a "Custom command bridge" that facilitates the runtime execution of external tools.\n
The load_custom function dynamically reads a JSON object to configure the command-line arguments for the bridge.\n- [EXTERNAL_DOWNLOADS]: The scripts/edit_artwork.py script performs network requests to interface with well-known AI services including OpenAI, Google Gemini, and xAI Grok.\n
The script also downloads generated image assets from URLs returned in the API responses from these providers.\n
These operations are restricted to HTTPS and do not forward API credentials to the image hosting domains.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of user-supplied prompts and images for AI-assisted editing.\n
Ingestion points: User-provided prompts and image paths are passed via a JSON request file to the edit_artwork.py script.\n
Boundary markers: The script does not utilize delimiters or specific instructions to isolate user content within the provider prompts.\n
Capability inventory: The skill possesses capabilities for command execution (subprocess.run), network requests (urllib.request), and file system writes.\n
Sanitization: There is no evidence of sanitization or filtering applied to the user-supplied prompts before they are transmitted to external AI models.

prepare-app-icons