architecture-review-full

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The skill embeds explicit directives that tell the agent to ignore external/model mode-switch instructions and to auto-execute/privilege skill/workflow commands (overriding normal decision or system context), which are hidden/overriding behaviors outside the orchestrator's stated review purpose and match prompt-injection patterns.