flame-audit

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

• Third-party content exposure detected (high risk: 0.75). The required workflow explicitly ingests runtime performance/trace data and network request details derived from the user-supplied URL (public web content), which can include outsider-authored free text (e.g., page/script/stack labels and any tool-returned summaries) into the LLM context via the trace/insights responses.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill includes explicit runtime commands that fetch and execute remote packages (e.g., npx -y chrome-devtools-mcp@latest and npx lighthouse --only-categories=performance --output=json --quiet), which would download and run third‑party code during skill execution.