The Agent Skills Directory

[Data Exposure & Exfiltration] (LOW): The skill contains multiple curl commands directed at tavusapi.com. While this domain is specific to the service and no sensitive local data is accessed, the domain is not on the predefined whitelist for network operations.
[Indirect Prompt Injection] (LOW): The skill accepts untrusted external data in the form of URLs (train_video_url, consent_video_url, callback_url).
Ingestion points: API request body parameters in SKILL.md.
Boundary markers: None (the data is interpolated directly into JSON payloads).
Capability inventory: Limited to curl network requests to the API endpoint.
Sanitization: None specified in the documentation templates.
[Prompt Injection] (SAFE): No attempts to override system prompts or bypass safety guidelines were detected.
[Credentials Unsafe] (SAFE): The skill correctly uses the placeholder YOUR_API_KEY instead of hardcoding actual secrets.

tavus-replica