The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text prompts for image generation and model training, which is an inherent feature of AI model integration.
Ingestion points: User input via GenerateRequest and FluxDevInput in SKILL.md.
Boundary markers: None; prompts are passed directly to the replicate client.
Capability inventory: The skill uses the Replicate API to generate images and manage model predictions.
Sanitization: No specific sanitization is performed on the input prompts, relying on the model's own safety checkers (which are enabled by default in the provided code).
[EXTERNAL_DOWNLOADS]: The skill references standard Python libraries to be installed from the official registry.
Evidence: Instructions to install replicate, httpx, python-dotenv, and pydantic. These are well-known packages for interacting with APIs and managing configurations.
[COMMAND_EXECUTION]: The documentation includes shell commands for local environment setup.
Evidence: pip install commands and .env configuration steps are documented to guide the user in setting up the integration. These commands do not involve privilege escalation or suspicious execution patterns.

replicate-integration