The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external content from research papers and retrieved web sources using Read, Grep, and WebFetch. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the audited text or the source material retrieved online.
Ingestion points: The skill reads local files via Read, Grep, and Glob, and fetches external web content via WebFetch (referenced in SKILL.md).
Boundary markers: There are no explicit instructions to the agent to treat processed data as untrusted or to ignore embedded instructions within the documents.
Capability inventory: The agent has access to WebSearch, WebFetch, and Write tools, which could be leveraged if a prompt injection attack is successful.
Sanitization: The instructions do not define any sanitization or validation logic for the external content before it is processed by the agent.

citation-audit