The Agent Skills Directory

Data Exposure & Exfiltration (LOW): The skill accesses the MAX_API_KEY environment variable and transmits it to internal.infquest.com. While this is consistent with the skill's stated purpose of authenticating with a video generation API, the domain is not on the pre-approved whitelist.
Indirect Prompt Injection (LOW): The skill possesses an attack surface for indirect prompt injection.
Ingestion points: User-provided prompts and image paths collected via the AskUserQuestion tool.
Boundary markers: Absent; user input is directly interpolated into the script execution command and API payload.
Capability inventory: The skill executes local Python scripts via uv run, writes files to the local disk, and performs network POST/GET requests.
Sanitization: No explicit sanitization or validation of the prompt string is performed before it is passed to the video generation model or used in the shell command.
Command Execution (SAFE): The skill executes a local Python script (video-gen.py). The script uses argparse and pathlib, which provides some protection against basic path traversal or command injection compared to raw string concatenation, though the agent-side invocation open "OUTPUT_PATH" relies on the agent's ability to handle potentially malicious paths safely.

video-gen