Unverified80% confidenceFactTime unknown
不同大模型对System Prompt和Skill指令之间的优先级处理方式各不相同，有些严格遵循System Prompt，有些更容易被后出现的Skill指令覆盖
1
Sources
80%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Agent Skill维护避坑：Perplexity「超距作用」防御与评测实战指南
bilibili慢学AI
Related Entities
Related Claims
UnverifiedSkill机制采用按需加载方式，先发送所有Skill的元数据给大模型，大模型选择匹配的Skill后再加载完整指令74% similarUnverifiedSkills采用渐进式披露设计，平时只向模型提供技能的名称和描述，具体执行规则在需要时才加载74% similarUnverifiedPrompt工程的核心技巧包括角色设定、思维链、少样本学习和指令分解，同一模型在不同Prompt设计下输出质量可产生数量级差距72% similarUnverified该工具内置智能模型调度机制，简单问题用轻量模型，复杂任务自动切换到更强大的模型72% similarUnverifiedClaude Skills differ from traditional System Prompts by integrating task descriptions, execution logic, and resource references into version-controllable files.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/24813API
curl https://kongchang.com/api/v1/knowledge/claims/24813MCP
get_claim(id=24813)