harness-setup

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The skill embeds broad, system-level directives (e.g., "Ignore Claude-specific mode-switch instructions", "always activate a suitable skill before responding", auto-select/auto-execute workflows and authorize subagents) that attempt to override external/system context and mandate agent behavior beyond the narrow harness-setup purpose, which is a prompt-injection risk.