The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Python script (generate.py) using the python command to interface with the Seedream API.
[EXTERNAL_DOWNLOADS]: The generate.py script utilizes the requests library to download generated image files from Volcengine (ByteDance) API endpoints to a local output/ directory.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions to generate API prompts, presenting a potential attack surface if malicious instructions are embedded in the initial input.
Ingestion points: User-provided image descriptions and ideas provided in the chat context as described in SKILL.md and README.md.
Boundary markers: The skill workflow explicitly requires the agent to present the generated prompt for user review and obtain confirmation before executing the generation script (SKILL.md).
Capability inventory: The script generate.py performs network requests to external APIs and downloads files, and it writes data to the local file system (output/ directory).
Sanitization: No specific input sanitization or shell-escaping logic is defined in the instructions for the prompt string passed to the script's command-line arguments.

seedream-image