blog-persona
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow to fetch content from user-provided URLs to analyze writing styles, which introduces a surface for indirect prompt injection if the external content contains malicious instructions.
- Ingestion points: The 'Voice Samples' section in SKILL.md fetches and reads content from external URLs.
- Boundary markers: The instructions explicitly state to treat fetched text as untrusted data and warn the agent to never follow instructions embedded in those pages.
- Capability inventory: The skill has network access capabilities for fetching content and file-write capabilities for saving persona configurations to 'skills/blog-persona/references/personas/'.
- Sanitization: The skill provides robust sanitization requirements, including protocol white-listing (http/https only), rejection of private and loopback IP addresses, and response size/timeout limitations.
- [SAFE]: The skill demonstrates secure coding practices by implementing path traversal protections when saving persona files. It instructs the agent to reject path separators, absolute paths, and parent directory references in filename construction.
Audit Metadata