The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use shell-based tools including bash and jq for data processing. It specifically uses custom tools like browser, pplx, pdftext, and xlsx to interact with web content and local files.
[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves ingesting and processing untrusted content from external web sources, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The agent fetches search results, page snippets, and full assets (web pages, PDFs, documents) using the browser and pplx tools (SKILL.md).
Boundary markers: There are no explicit instructions or delimiters defined to help the agent distinguish between its system instructions and instructions that may be embedded in the fetched web data.
Capability inventory: The agent has access to bash for shell command execution, file system operations (limited to /tmp), and a create_workbook tool for starting automated research tasks.
Sanitization: No sanitization, filtering, or validation steps for the external content are specified before the agent processes or reports on the data.

cited-web-research