Insight for: claude code 用kiro反代不吃pua

This issue reveals a fundamental challenge in controlling AI agent behavior, specifically the 'pua' skill's inability to influence Claude Code when routed through a 'kiro' proxy. The 'I can't discuss that' response indicates a potential content filter or policy enforcement, possibly triggered by the proxy or the 'pua' rhetoric itself. This highlights a critical friction point between desired agent manipulation and underlying platform guardrails. Market implications include limitations on advanced agent control strategies, particularly those pushing ethical or behavioral boundaries, suggesting platform providers are actively restricting certain types of agent interaction.