blog-brand
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting user-provided brand data and existing persona JSON to generate documentation files intended for consumption by other agents.
- Ingestion points: Interactive interview responses and data from the
blog-personaJSON file. - Boundary markers: The generated output uses Markdown headers and lists to provide structure between different data fields.
- Capability inventory: File system write access to the project root for creating
BRAND.mdandVOICE.md. - Sanitization: Includes basic validation for answer length to ensure content quality, though it does not explicitly define sanitization or escaping for prompt-sensitive characters.
- [SAFE]: The skill performs legitimate project management and configuration tasks. Its file operations are limited to the project root and are essential for its primary function of maintaining editorial context.
Audit Metadata