openclaw-autoreview

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

• Third-party content exposure detected (high risk: 0.75). The runtime LLM context is built from the local git change bundle plus optional --prompt-file/--dataset contents (repo-relative files read as UTF-8 prose), which can include outsider-authored free text from the reviewed branch/PR; e.g., branch_bundle()/commit_bundle()/local_bundle() include git diff/show output and validate_evidence_file() reads arbitrary repo files into the prompt.