skill-creator

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

• Third-party content exposure detected (medium risk: 0.30). In the eval viewer runtime, eval-viewer/generate_review.py recursively reads text from previously-run outputs/ files (and transcript.md fallback prompt) and embeds it into the served HTML, so an outsider who can submit text that becomes an output/transcript file can indirectly influence LLM-ingested free text at review time.

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

• Potentially malicious external URL detected (high risk: 0.70). The viewer HTML loads and will execute remote JavaScript at runtime — specifically the SheetJS script https://cdn.sheetjs.com/xlsx-0.20.3/package/dist/xlsx.full.min.js which the viewer relies on to render .xlsx files in-browser.