The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from untrusted external sources, including web pages, YouTube transcripts, and various document formats (PDF, DOCX, PPTX) via the researcher and context7 MCP servers.
Ingestion points: Untrusted data enters the agent's context through tools such as mcp__researcher__search_and_scrape, mcp__researcher__scrape_page, and documentation queries specified in SKILL.md.
Boundary markers: The instructions do not define boundary markers or use specific delimiters to separate the retrieved external content from the agent's internal instruction set.
Capability inventory: While the skill itself focuses on reading data, the processed content could contain instructions targeting other capabilities available to the agent (e.g., file system access or network operations) if they exist in the broader environment.
Sanitization: There are no instructions provided for sanitizing, filtering, or validating the content retrieved from external sources before it is analyzed by the agent.

sage