dependency-confusion

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). Yes — the list includes attacker-controlled callback endpoints used in dependency-confusion PoCs and a public GitHub PoC repository that can be used to publish packages with lifecycle hooks, which are common vectors to execute malicious code during installs.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document is a high-risk, intentional supply-chain attack playbook: it instructs publishing squatted packages with install-time hooks that execute remote callbacks, enabling remote code execution and data exfiltration on victim build/developer hosts.

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 explicitly instructs publishing packages with lifecycle hooks (npm pre/postinstall and pip install hooks) that execute callbacks/commands on victim machines and tells the reader to register/publish such packages, which is direct, actionable guidance to execute code and thereby modify/compromise the state of systems.