The Agent Skills Directory

[SAFE]: No malicious behavior or high-risk patterns were detected. The skill leverages a local CLI tool ('renku') for its operations, which is a secure way to interface with external APIs without exposing secrets to the agent's memory or persistent storage.
[SAFE]: The skill enforces a segregated file structure for temporary operations, media, and quality assurance evidence within a dedicated 'tmp/' tree. This practice prevents unauthorized access to or modification of files in the project's root directory.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from local reference files and model guides to configure generation requests. This represents a potential surface for indirect prompt injection, although the risk is low as the data source is internal to the skill's project structure.
Ingestion points: Metadata and guides are read from 'references/supported-routes.json' and 'references/model-guides/' (SKILL.md).
Boundary markers: No explicit markers are used to isolate these values when they are interpolated into the agent's planning or command construction.
Capability inventory: The agent is authorized to execute 'renku' CLI commands using inputs derived from these files.
Sanitization: There is no explicit sanitization of the content found within the model guides before it is used by the agent to author requests.

elevenlabs-media-provider