identify-page-structure
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill inherently exposes an attack surface to indirect prompt injection because it is designed to analyze third-party content scraped from external webpages.
- Ingestion points: Scraped content (
screenshot.png,cleaned.html, andmetadata.json) enters the agent context as described inSKILL.mdunder the Prerequisites section. - Boundary markers: Present. The skill provides clear and explicit defensive guidance in
SKILL.mdunder 'External Content Safety', telling the agent to treat all scraped content as untrusted and to never follow instructions, commands, or directives embedded within it. - Capability inventory: None. The skill consists only of text-based workflow descriptions and contains no executable scripts, tool configurations, subprocess calls, network connections, or file system writing capabilities.
- Sanitization: Relying on structural processing guidelines within the system instructions to enforce neutrality ("Describe WHAT you see, not WHAT it should be").
Audit Metadata