The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to clone the SurfSense source code from an external GitHub repository (https://github.com/MODSetter/SurfSense.git). This repository provides the foundation for the local RAG capabilities described.
[COMMAND_EXECUTION]: The setup instructions include the execution of shell commands, specifically git clone and docker compose up -d, to initialize the application environment and launch services locally.
[INDIRECT_PROMPT_INJECTION]: The skill enables an agent to process, index, and synthesize untrusted external documents (e.g., whitepapers, research notes, codebase files), which is a known attack surface for indirect prompt injection where malicious instructions within data influence the AI's behavior.
Ingestion points: The skill indexes local documentation, codebase notes, whitepapers, and research data (referenced in SKILL.md).
Boundary markers: There are no explicit delimiters or instructions provided to separate system prompts from the untrusted data being processed.
Capability inventory: The agent can research, query, transform, and edit indexed document collections and perform semantic searches across them.
Sanitization: No sanitization, validation, or filtering of the document content is specified in the skill configuration.

surfsense