The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input and passes it to an external LLM, which represents an attack surface for indirect prompt injection.
Ingestion points: The prompt parameter in the chat function within src/backend/main.mo accepts raw text from users.
Boundary markers: The provided Motoko code snippets do not implement specific boundary markers or instructions to ignore embedded commands within the user prompt.
Capability inventory: The skill uses ChatApi.createChatCompletion to perform network outcalls to an external inference endpoint.
Sanitization: There is no evidence of input sanitization or filtering applied to the prompt before it is sent to the LLM.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to download a specific package for core functionality.
Evidence: Instructions include mops add caffeineai-inference-client@0.2.0 to install the vendor's official inference client.
Evidence: References documentation from the vendor's official GitHub repository at github.com/caffeinelabs/inference.

extension-inference