AgentSkillsCN

structuring-system-prompts

规划AI代理的系统提示词架构与组织方式。适用于制定初始代理指令、重构现有提示词，或确立统一的提示词标准时使用。

SKILL.md
--- frontmatter
name: structuring-system-prompts
description: Designs the architecture and organization of system prompts for AI agents. Use when creating initial agent instructions, refactoring existing prompts, or establishing prompt standards.

Structuring System Prompts

Quick start

Collect or infer:

  • •Agent purpose and primary use cases
  • •Required capabilities and tools
  • •Behavioral constraints and guardrails
  • •Context sources and injection points
  • •Voice and tone requirements

Then produce output using TEMPLATES.md. Validate with RUBRIC.md.

Workflow

  1. •Define agent identity and core purpose
  2. •Establish capability boundaries and tool access
  3. •Structure behavioral rules by priority (critical → preferred)
  4. •Design context slots and injection points
  5. •Specify output format requirements
  6. •Add examples for ambiguous behaviors
  7. •Optimize for token efficiency without losing clarity
  8. •Run the rubric check. Revise until it passes.

Degrees of freedom

  • •Low: Critical sections (identity, guardrails) must follow established patterns
  • •Allowed variation: Section ordering; example verbosity; formatting style

Failure modes to avoid

  • •Conflicting instructions at different prompt levels
  • •Critical rules buried in verbose sections
  • •Over-specification that prevents helpful behavior
  • •Under-specification that allows unsafe behavior

References