The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides an extensive interface for processing untrusted external data through various machine learning pipelines.
Ingestion points: Multiple entry points for external data are documented in references/detailed-guide.md, including classifier() for text and images, and transcriber() for audio files.
Boundary markers: The provided documentation and examples do not include delimiters or specific instructions to help the agent distinguish between its core logic and instructions potentially embedded in the processed data.
Capability inventory: The skill documentation reveals significant system capabilities, such as file system writes (fs.writeFile in references/EXAMPLES.md) and configurable network operations (env.fetch in references/CONFIGURATION.md).
Sanitization: There is no evidence of sanitization or validation of the model outputs before they are processed by the agent, which could allow maliciously crafted data to influence the agent's behavior.
[EXTERNAL_DOWNLOADS]: The skill fetches code and model assets from external sources.
Downloads JavaScript modules from JSDelivr, an established content delivery network for web developers.
Retrieves machine learning model weights and configurations from the Hugging Face Hub, an established repository for AI assets.

transformers-js