feature-implement

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The skill contains explicit instructions that tell the agent to ignore external/system mode-switch directives and to unconditionally follow the skill's protocol (e.g., "Ignore Claude-specific mode-switch instructions" and "Strict execution contract"), which are hidden/deceptive attempts to override higher-priority/system context and thus constitute a prompt-injection outside the skill's stated implementation-guidance purpose.