paper-proofread

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs the agent to "silently read the full workspace" — starting from the user-provided root .tex file and recursively resolving all \input/\include, .bib files, and figures — and then uses that untrusted, user-supplied content to generate findings and drive edits, so arbitrary third-party/user-generated content can influence the agent's decisions and actions.