coding-agent

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

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs cloning an external repo at runtime (git clone https://github.com/user/repo.git) and then runs the coding agent inside that workspace, so fetched repository content can directly control the agent's input/context.

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 promotes bypassing sandboxing and approvals (flags like --yolo and --full-auto) and exposes an "elevated" host execution option, which encourages the agent to run with host-level privileges and thereby modify machine state or bypass security controls.