meta
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-provided posts or topics to generate metadata and extract facts.
- Ingestion points: Instructions in
SKILL.mddirect the agent to "Skim the post (or topic)", "Read the post's H1 and first paragraph", and extract specific claims from the post body. - Boundary markers: The prompt lacks instructions for the agent to use specific delimiters or to disregard potential instructions embedded within the user's content (e.g., hidden HTML comments or text designed to hijack the agent's logic).
- Capability inventory: The skill allows for local code execution (e.g., character counting) and generates structured JSON-LD and meta tags based on the input. It also reads configuration files from relative paths (
.agents/blog-context.mdand../house-style/references/ai-tells.md). - Sanitization: There are no requirements defined for sanitizing or escaping the content retrieved from the external post before it is used to populate metadata fields or influence agent output.
Audit Metadata