The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a Retrieval-Augmented Generation (RAG) pattern that processes untrusted data from external web sources and injects it directly into LLM prompts.
Ingestion points: Data enters the agent's context through tavily/search-assistant, exa/search, and tavily/extract outputs as demonstrated in SKILL.md.
Boundary markers: Prompt templates lack explicit delimiters or instructions to the LLM to ignore potentially malicious commands embedded in the retrieved web content.
Capability inventory: The skill is configured with Bash(belt *) permissions, allowing it to execute tools on the platform.
Sanitization: No evidence of data sanitization or filtering of the retrieved content before prompt interpolation.
[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to download and install external software and configurations from the vendor's infrastructure.
Evidence: References installation of the belt-sh/cli and various skill packages from the inference-sh organization via npx.
Evidence: Links to installation instructions hosted at https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.

ai-rag-pipeline