The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from an external website (epfl.ch), which represents a surface for indirect prompt injection attacks.
Ingestion points: The epfl_ch__render_page and epfl_ch__read_page tools described in SKILL.md retrieve content from external web pages.
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands within the retrieved text.
Capability inventory: The skill has the capability to perform network requests to the vendor's API to fetch site data.
Sanitization: The Unbrowse service processes the raw site data into "readable text and links" before returning it to the agent.
[EXTERNAL_DOWNLOADS]: The skill configuration and instructions reference resources hosted on the vendor's domain.
Evidence: The MCP server endpoint is located at https://unbrowse.ai/mcp/epfl.ch as seen in README.md and server.json.
Evidence: Installation instructions in README.md use npx skills add pointing to the vendor's website.
Context: All external references are restricted to the official infrastructure of the skill author (unbrowse-ai).

epfl-ch