The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements Retrieval Augmented Generation (RAG) workflows that ingest data from external sources and interpolate it into LLM prompts, creating an indirect prompt injection surface.
Ingestion points: The skill captures output from tools like tavily/search-assistant, exa/search, and tavily/extract into shell variables (e.g., $SEARCH, $CONTENT, $EVIDENCE) which are then placed directly into prompt strings.
Boundary markers: Prompt templates use basic text labels such as "Search Results:" or "Source 1 (Tavily):" but lack strong structural delimiters or specific instructions to the model to ignore potential directives embedded within the retrieved data.
Capability inventory: The skill is granted access to the Bash tool to run belt commands, which allows for network communication and task execution.
Sanitization: No explicit content filtering, escaping, or validation is performed on the retrieved content before it is processed by the LLM.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions to download the platform's CLI tool and suggests adding specialized modules from the provider's repositories.
The skill suggests installing belt-sh/cli and several inference-sh/skills modules (e.g., web-search, llm-models, ai-content-pipeline) via the npx skills add command.
It links to installation guidelines hosted on GitHub: https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.
[COMMAND_EXECUTION]: The skill relies on the Bash tool to execute belt CLI commands, which is the primary mechanism for its search, extraction, and inference operations.

ai-rag-pipeline