canonical-header

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines an audit process that ingests external data including rendered HTML, structured data, and HTTP response headers (ingestion points in SKILL.md and references/rule.md). While this surface could theoretically be used for indirect prompt injection if the audited codebases contain malicious directives, the risk is negligible as the skill lacks automated execution tools, file-system write capabilities, or network communication scripts. No specific boundary markers or data sanitization routines are implemented for the external content reviewed.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 05:58 PM
Security Audit — agent-trust-hub — canonical-header