codexcont-middleware

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). One URL (https://chatgpt.com/backend-api/codex/responses) is suspicious because chatgpt.com is not the official OpenAI API host and can be a typosquatting/third‑party upstream that might intercept credentials or deliver untrusted binaries; the other URLs are official/internal or not download links.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The documentation exposes features that can be intentionally abused for credential theft and data exfiltration (per-request upstream override, auth passthrough, and a suspicious default upstream) and retains sensitive conversation state in-process, indicating a high risk of deliberate misuse or supply-chain/backdoor behavior if deployed or configured maliciously.

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

• Potentially malicious external URL detected (high risk: 1.00). The README instructs cloning and then running code from the remote repository https://github.com/neteroster/CodexCont.git (git clone ... then uv run python run.py), which fetches remote code that is executed locally as part of setup/runtime.