source-investigator

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

• Third-party content exposure detected (high risk: 0.85). The workflow clones an exact external repository into .tmp/repos/ and then uses subagents to inspect it, meaning outsider-authored source code (from the cloned repo) is read as text at runtime and can be included in the agent’s LLM context via the subagents’ findings.