deepagents-thread-inspector

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow runs scripts/inspect_sessions.py, which reads an on-disk SQLite sessions DB at runtime and reconstructs/stages stored checkpoint/writes “messages” (outsider-originating free text may be present in local state) into JSON printed to stdout for the agent to consume into its LLM context.