dream

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

• Third-party content exposure detected (high risk: 0.75). Outsider-authored free text is ingested because Step 2 fetches project memories via get_memories(...) and Step 3/4/5 operate on their text content; those memories can include messages authored by other parties (e.g., comments/issues/wiki text) that the user did not author, and the skill then feeds that content into the agent’s LLM context for analysis and diff generation.