The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is susceptible to indirect prompt injection because it ingests untrusted content from documents to identify factual claims without implementing strict boundary markers.
Ingestion points: The agent scans user-provided documents (SKILL.md, Step 1) to identify claims for verification.
Boundary markers: Absent; there are no instructions to use delimiters or ignore embedded instructions within the document content being analyzed.
Capability inventory: The skill utilizes an Edit tool for file modification and web search for verification.
Sanitization: Absent; the skill does not explicitly sanitize or validate extracted claims before processing them for search or modification.
[Data Exposure & Exfiltration] (LOW): Information extracted from documents is used to generate web search queries to external search engines.
Evidence: If a document contains sensitive internal information that is incorrectly identified as a verifiable claim, that information would be transmitted to external search providers. This is a functional risk rather than a malicious exfiltration attempt.

fact-checker