research

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

• Third-party content exposure detected (medium risk: 0.30). The market-pulse mode requires pulling data from sources like Reddit, Hacker News, X/social, and the open web without the agent first selecting a specific item, which allows outsiders to submit free text into monitored feeds.

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

• Potentially malicious external URL detected (medium risk: 0.30). The repository locator referenced in README.md (harshsinghmp/muse-skills) is an ordinary third-party repository without strong official provenance or authentication guarantees, constituting a common/reputable unofficial operational dependency.