faq-page
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data and generate code-based artifacts, creating a potential surface for indirect injection attacks.
- Ingestion points: The instructions in
SKILL.mddirect the agent to read an activeDESIGN.mdfile and extract domain-specific content from a user-supplied "brief". - Boundary markers: The workflow lacks instructions for using delimiters or explicit safety prompts to ignore potentially malicious instructions embedded within the ingested data.
- Capability inventory: The skill produces a single HTML document with inline CSS and JavaScript (
example.html) that implements search and filtering logic. - Sanitization: There are no explicit requirements in the workflow to sanitize, escape, or validate input from the user brief before interpolating it into the final HTML/JS output.
Audit Metadata