edit-tool

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The document exposes multiple powerful features that can be abused as backdoors or for data exfiltration—notably bypassPermissions/dontAsk/acceptEdits modes, arbitrary hook/command execution, auto-enabled Read/Edit/Write for persistent memory, background agents that continue on permission failures, inline mcpServers (arbitrary npx/commands), script execution, and full-skill injection into agent contexts—together these enable remote code execution, credential harvesting, persistent exfiltration, and supply-chain style abuse if misused.