rde-eval

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can enter the LLM context via the cloud git:https://github.com/...@<ref> / git:https://github.com/...@<branch> runtime path: the harness clones third-party OSS repos and then reads their source files (e.g., file contents / diagnostics context) as readable text for analysis, which the agent/LLM may incorporate into prompts (indirect prompt injection risk from repo-authored text).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The harness's cloud mode explicitly accepts and uses the git spec git:https://github.com/millionco/react-doctor@ at runtime to clone and build the remote monorepo inside Vercel sandboxes (i.e., fetching and executing remote code), so this URL is a required runtime dependency that executes external code.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)