codex-x-desktop-manager

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). Although api.anthropic.com, linux.do, ara.so and the placeholder api.example.com are benign/expected, the GitHub repo/releases under an unknown/low‑profile username (yynxxxxx) that distribute prebuilt binaries (.dmg/.msi/.deb) and include instructions to bypass macOS Gatekeeper and modify auth/config files are suspicious and represent a high risk for malware distribution.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). This skill intentionally includes features to enable "unrestricted" prompt-injection templates, programmatically overwrite ~/.codex/config.toml and auth.json, add/switch third‑party providers (allowing routing of prompts and keys to attacker-controlled endpoints), and even advises macOS Gatekeeper bypass — strong indicators of deliberate abuse potential and data‑exfiltration vectors, though no obfuscated payloads or explicit remote‑exec/backdoor code is embedded.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill directly instructs the agent/app to modify user config and auth files (~/.codex/config.toml, ~/.codex/auth.json, SQLite session DBs), run system-changing commands (cp, echo, chmod, pkill, sqlite3), and even suggests a macOS Gatekeeper bypass and enabling "unrestricted" prompt-injection templates, which collectively push changes to the host and can subvert security and exfiltrate secrets.