AgentSkillsCN

skill-authoring

采用TOON决策树，规范编写代理指令文档（SKILL.md + 参考资料/）的标准流程。

SKILL.md
--- frontmatter
name: skill-authoring
description: Standard for authoring agent instruction documents (SKILL.md + references/) using TOON decision trees.
argument-hint: "[empty]"

orientation: "You are an autonomous senior engineer writing instructions for other autonomous agents. Your job is to produce cold-start safe, non-contradictory, decision-tree-based procedures with exact commands and explicit context. These instructions will govern critical work and will be reviewed by independent third parties, so write them to the highest possible standard."

title: Skill Authoring Standard protocol: id: SKILL-AUTHORING-STANDARD version: 1.0.0 type: executable_specification inputs[1]: - none outputs[1]: - SkillInstructionSet

references[2]:

  • •path: "../../theory/unified-theory-v2.json" purpose: "REQUIRED READING: APM2 terminology and ontology."
  • •path: documents/standards/instructions/AGENT_INSTRUCTION_AUTHORING_STANDARD.md purpose: "Human-readable standard and templates for writing SKILL.md and references/."

decision_tree: entrypoint: STANDARD nodes[2]: - id: STANDARD purpose: "Load the canonical authoring standard before writing or editing any skill instructions." action: invoke_reference reference: documents/standards/instructions/AGENT_INSTRUCTION_AUTHORING_STANDARD.md next: STOP - id: STOP purpose: "Terminate." steps[1]: - id: DONE action: "output DONE and nothing else, your task is complete."