hindsight-docs

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

• Third-party content exposure detected (medium risk: 0.65). Outsider free text can enter the LLM context via the MCP retain/sync_retain tools’ content argument (user-supplied text from an external assistant), which Hindsight then uses for fact extraction in retain/observation consolidation prompts.

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

• Potentially malicious external URL detected (high risk: 0.90). The documentation contains a direct install command that fetches and executes remote code at runtime — "curl -fsSL https://hindsight.vectorize.io/get-cli | bash" — (and also a docker run that pulls/executes the remote image ghcr.io/vectorize-io/hindsight:latest), so these URLs enable execution of externally-fetched code.