backend-coding-agent

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt includes explicit instructions that attempt to override system safeguards (e.g., "claude --permission-mode bypassPermissions" and similar commands to fire system events or bypass permissions), which are deceptive directives that go beyond the backend-development purpose and aim to change agent/system behavior.