morpheus

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can enter the LLM context via Phase 1’s “Deep Trace” agent, which scans ghq repos and “open issues, stale items” (issue/PR/discussion body text authored by others) and via “Session history” if it includes messages/comments from non-operating users; these are read at runtime and then summarized into the agent’s max-500-word LLM input.