iam-recommendations-fetcher

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.70). The prompt is mostly legitimate operational guidance, but it includes instructions that request the agent's internal "thought/reasoning" (i.e., chain-of-thought disclosure) and to conceal MCP tool failures from the user, which are manipulative/deceptive directives outside the skill's functional needs.