herdr-agent-comms

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

• Third-party content exposure detected (high risk: 0.78). Risk via agent transcript scraping: during required Phase 5/3 read-after-wait, the workflow calls herdr pane read ... --source recent-unwrapped and then relays the resulting free-form text into the skill/agent context (and the “reply delta” itself can contain prompt-injection text authored by the outsider agent/session).