compiler

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow ingests outsider-authored free text when it “READ all inputs” and, for inputs that are URLs or repos, it fetches/clones public web pages or third-party repositories and then renders/reads their contents (including arbitrary text from those sources) into the agent’s LLM context.