competitor-watch

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

• Third-party content exposure detected (medium risk: 0.30). The competitor-watch runtime “Gather” step captures text from already-configured watched public URLs/selectors (e.g., CSS/XPath/JSONPath targets) and may summarize/diff the retrieved page content, so outsider-authored content on those pages is ingested and could contain indirect prompt-injection strings.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill's docker-compose example pulls and runs remote container images (ghcr.io/dgtlmoon/changedetection.io and dgtlmoon/sockpuppetbrowser:latest) which will be fetched and executed at runtime when bringing up the runnable changedetection.io stack.