The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it fetches and processes untrusted content from the web using WebFetch. This could allow an attacker to place malicious instructions on a webpage that the agent might follow during research.
Ingestion points: The skill uses WebFetch to read the full content of pages discovered via WebSearch (Phase 2, Step 5).
Boundary markers: There are no instructions to wrap fetched external content in delimiters or to include 'ignore embedded instructions' warnings when the agent processes this data.
Capability inventory: The skill utilizes powerful tools including WebSearch and WebFetch for network access, alongside Grep and Glob for searching the local codebase (Phase 2, Step 6).
Sanitization: The instructions do not specify any sanitization, validation, or filtering of the content retrieved from external URLs before it is analyzed and synthesized into the final report.

researcher