The Agent Skills Directory

[SAFE]: The skill uses environment variables (REPLICATE_API_TOKEN) for handling credentials, ensuring that sensitive tokens are not hardcoded within the instructions.- [SAFE]: All external requests are directed to the official Replicate API infrastructure (api.replicate.com), which is a well-known and trusted service provider.- [SAFE]: No obfuscation, remote code execution patterns, or unauthorized persistence mechanisms were identified in the skill content.- [SAFE]: Indirect Prompt Injection Surface: The skill processes untrusted user prompts to interact with external models (SKILL.md). 1. Ingestion points: input.prompt field in the fetch body. 2. Boundary markers: Absent. 3. Capability inventory: Performs network POST requests via fetch. 4. Sanitization: None provided in example code. This surface is standard for AI generation tools and does not indicate malicious intent.

replicate