skills/howells/arc/responsive/Gen Agent Trust Hub

responsive

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's primary functions—reading design documents, discovering routes via package.json, and automating browser screenshots—are consistent with its stated purpose of responsive design auditing.
  • [DATA_EXPOSURE_&_EXFILTRATION]: Network activity is restricted to a user-provided local development server URL. There is no evidence of the skill accessing sensitive system files (like SSH keys or credentials) or transmitting data to external domains.
  • [COMMAND_EXECUTION]: Command execution is limited to standard git operations (git add, git commit) for versioning the applied fixes. No unauthorized or dangerous shell commands were found.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from project files (design docs, package.json) and visual content from the website it audits. While this presents a standard surface for indirect prompt injection, the skill includes human-in-the-loop checkpoints (using AskUserQuestion) for route confirmation and fix validation, which mitigates the risk of autonomous malicious actions.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 01:58 PM