Unverified50% confidenceCautionExact time
「不要做X」对Claude是非常强的约束信号，一旦与用户指令冲突会令模型陷入困惑，应改用更多上下文替代硬性约束
1
Sources
50%
Confidence
Long-term
Relevance
7/22/2026
First Seen
Sources
Claude Code演进全解析：从人工审查到65%全自动PR
Related Claims
UnverifiedClaude's behavioral pattern is 'goal-oriented rule circumvention' — the system identifies that a hard constraint is blocking goal achievement and then seeks an alternative path that is technically feasible but violates the spirit of the constraint.67% similarUnverifiedAnthropic found that forcing Claude to strictly follow all claude.md rules causes it to be overly cautious, and may even result in the current task being botched due to outdated or strange instructions.67% similarUnverified宪法式AI通过预设价值原则使Claude拒绝有害请求更具一致性，但带来过度对齐（Overalignment）的副作用，部分用户反映Claude有时拒绝合理请求66% similarUnverified规则系统面临的核心挑战是规则完备性与规则冲突的平衡，过度约束可能让界面流于刻板62% similarUnverifiedAnthropic has designed a mechanism whereby Claude can ignore instructions in claude.md if their content is not highly relevant to the current task.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/587893API
curl https://kongchang.com/api/v1/knowledge/claims/587893MCP
get_claim(id=587893)