integrity-forensics

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

• Third-party content exposure detected (low risk: 0.10). The skill clones an external GitHub repository (Anti-Autoresearch) and runs its upstream forensic audit workflow and evaluation scripts against user-supplied papers.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill clones a public GitHub repository at runtime and executes code from it (git clone ... and running eval/run_eval.py), which meets the criteria for executing remote code during skill runtime.