The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a pattern that accepts untrusted data from users and processes it using an LLM.
Ingestion points: The chat function in src/backend/main.mo accepts a prompt : Text parameter directly from the caller.
Boundary markers: The provided code examples do not include delimiters or instructions for the LLM to ignore potentially malicious embedded content within the prompt.
Capability inventory: The skill uses ChatApi.createChatCompletion in src/backend/lib/inference.mo to perform network operations (HTTP outcalls) to an external inference endpoint.
Sanitization: There is no evidence of input validation, escaping, or filtering of the prompt variable before it is interpolated into the ChatCompletionRequest.
[EXTERNAL_DOWNLOADS]: The skill relies on external libraries and source code hosted on remote servers.
The instructions specify adding the caffeineai-inference-client package via the Mops package manager (mops add caffeineai-inference-client@0.1.0).
The skill references the official GitHub repository for the inference library at https://github.com/caffeinelabs/inference/releases/tag/public-api-v0.1.0.

extension-inference