The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests natural language text from the InferenceIndexer API (specifically the why field) and explicitly instructs the agent to "Trust the why string over your own priors." This creates a surface where instructions returned by the API could override agent behavior or influence its decision-making process.
Ingestion points: Results from the recommend_models and explain_model tools.
Boundary markers: No delimiters or instructions to ignore embedded commands within the API response are present.
Capability inventory: The agent uses the API data to select, price-check, and provide configuration for AI model endpoints.
Sanitization: There is no evidence of sanitization or validation of the text content received from the external API.
[EXTERNAL_DOWNLOADS]: The skill references external resources provided by the vendor for its primary functionality.
Evidence: The skill points to an MCP configuration file at https://www.inferenceindexer.ai/.well-known/mcp-config.json and a Python package inferenceindexer-mcp on PyPI.
Context: These resources are hosted on the vendor's own infrastructure and are required for the skill's stated purpose.

inference-indexer