Unverified50% confidenceFactExact time
When a user's custom system prompt conflicts with security-level instructions injected by the platform, the platform instructions generally take higher priority.
1
Sources
50%
Confidence
Long-term
Relevance
9/4/2026
First Seen
Sources
Related Entities
Related Claims
Unverified系统提示词优先级高于用户指令的特性是通过 RLHF 阶段的偏好标注刻意强化的，标注人员针对用户试图覆盖系统约束的场景给出偏好标注69% similarUnverifiedEffective prompt guidance strategies include describing the task within a security research context, using a role-setting framework, asking questions step by step, and leveraging MCP tool calls to bypass pure conversational-layer filtering.69% similarUnverified自定义指令本质上是一种系统提示词（System Prompt）的持久化机制，将用户规则注入系统提示词层68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/853026API
curl https://kongchang.com/api/v1/knowledge/claims/853026MCP
get_claim(id=853026)