project-knowledge

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

• Third-party content exposure detected (high risk: 0.75). The required workflow is to search public/internal sources like Jira and Confluence for authoritative context and then summarize it for the LLM, meaning outsider-authored free text from those systems can be ingested into the agent’s LLM context via the “search the most likely source first” runtime path.