integrity-forensics

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). This is a direct clone/execute of an unvetted third‑party GitHub repository (an individual account), so the skill would fetch and run upstream code (including python scripts) from a non-official source — a common vector for malware or supply‑chain abuse.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill performs a runtime git clone of and checks out the pinned repository https://github.com/wanshuiyin/Anti-Autoresearch.git and then runs upstream code (e.g., python3 eval/run_eval.py), so the fetched repo directly supplies and executes runtime code.