company-brain

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

• Third-party content exposure detected (medium risk: 0.30). In query mode, the skill ingests free-text from all first-party vault files under meetings/, people/, companies/, etc. (which are populated from outsider-authored sources like Slack, email, sales call transcripts, or support tickets via auto-sync), so an outsider can inject poison by submitting content that gets captured into the vault and later compiled/queried without additional selection constraints beyond those files.