cli-audit

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). Yes: the required workflow fetches a user-supplied live page URL at runtime (via page.goto(targetUrl)), then ingests its extracted accessibility/DOM-derived content into the LLM context indirectly through the emitted JSON fields axe and accessibilityTree printed to stdout (JSON.stringify(report, ...)).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The SKILL.md requires running npx --package=github:mgifford/accessibility-skills (github:mgifford/accessibility-skills) at runtime, which fetches and executes remote code from that repository, so the skill depends on external code executed during runtime.