best-practices

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill recommends using Cloudflare's cdnjs as a vetted mirror for loading polyfills to mitigate supply-chain risks associated with other third-party services (SKILL.md).\n- [INDIRECT_PROMPT_INJECTION]: The skill's workflow for performing security and quality audits involves processing data from rendered web pages and browser logs, establishing a surface for indirect prompt injection (SKILL.md).\n
  • Ingestion points: The agent is instructed to ingest and analyze rendered page content, console logs, and network failures using tools like lighthouse_audit or the Lighthouse CLI (SKILL.md).\n
  • Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the audited web content.\n
  • Capability inventory: The skill enables reading external content via Lighthouse and browser tools, and provides workflows for the agent to modify the application's source code to fix identified issues (SKILL.md).\n
  • Sanitization: No explicit sanitization or filtering of the audit input (the external website content) is specified, allowing potential injection attacks to reach the agent's context.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 06:14 AM
Security Audit — agent-trust-hub — best-practices