web-quality-audit

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external sources and has the capability to perform actions based on that data.\n
  • Ingestion points: The skill ingests untrusted data from URLs, page source, screenshots, and repositories as defined in SKILL_CARD.md.\n
  • Boundary markers: The instructions lack explicit boundary markers or directives to isolate external content from system directives.\n
  • Capability inventory: The skill is authorized to modify local code files to fix issues and execute browser-based tools for auditing.\n
  • Sanitization: There is no documented procedure for sanitizing or validating ingested content before it is processed or used to inform actions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 03:33 PM
Security Audit — agent-trust-hub — web-quality-audit