一键导入
documentation
Human-facing documentation (README.md) conventions. Use when writing docs for human readers. For agent-facing docs, use /agent-docs instead.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Human-facing documentation (README.md) conventions. Use when writing docs for human readers. For agent-facing docs, use /agent-docs instead.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | documentation |
| description | Human-facing documentation (README.md) conventions. Use when writing docs for human readers. For agent-facing docs, use /agent-docs instead. |
| user-invocable | false |
Apply these conventions when writing documentation.
| Aspect | Human-Facing (README, guides) | Agent-Facing (CLAUDE.md) |
|---|---|---|
| Length | Clear over brief | Extremely terse |
| Grammar | Complete sentences | Optional; sacrifice for brevity |
| Tone | Friendly, approachable | Direct, operational |
| Context | Explain "why" not just "what" | No explanations needed |
| Examples | Help understanding | Only if essential for correct usage |
| Human-facing (README) | Agent-facing (CLAUDE.md) |
|---|---|
| "Run the test suite before submitting a pull request." | pnpm test before PR |
| "This package uses Effect for error handling and concurrency." | Effect for errors, concurrency |
README is the entry point for humans. Template:
# Directory Name
One-line description of what this contains.
## Overview
2-3 sentences expanding on purpose and scope.
## Structure
Brief explanation of how contents are organized.
CLAUDE.md tells agents how to work here—commands, patterns, constraints. Template:
# [Directory Name]
> See README.md for context.
## Agent Instructions
[Operational details for working in this directory]
| Content type | Example |
|---|---|
| Operational constraints | pnpm typecheck after .ts edits |
| Pattern preferences | Composition over inheritance |
| Tool-specific commands | nx affected not nx run-many |
| Warnings | Don't modify dist/ |
Required for guides and specs:
---
status: active
description: When to consult this document and what guidance it provides.
---
| Status | Reliability | Usage |
|---|---|---|
placeholder | Low | Do not rely on |
draft | Medium | Use directionally; subject to change |
active | High | Use as authoritative source |
deprecated | N/A | Seek superseding documents |
status: field with appropriate value from Status Values tabledescription: explaining when to consult this document| Principle | Description |
|---|---|
| Show before tell | Example early grounds abstract rules |
| Concept before exception | Define X before "when not to use X" |
| Progressive disclosure | Simple core first, nuanced details later |
| Inverted pyramid | Key takeaway first, details follow |
> See README.md for context.When linking to other documents:
Effect service definition, interface design, error types, retries. Use when creating new services or defining error hierarchies.
AXM - Agent Extension Manager: Use for any operation (install/create/new/edit/update/add/remove/delete/publish/find/discover) on agent skills, subagents, slash commands/stored prompts, MCP servers, context packages, rule extensions, hook extensions, or packs — e.g. "create a skill", "make a /command", "add a subagent", "build an MCP server", "publish an extension". Use this BEFORE hand-authoring or editing any SKILL.md, slash-command, subagent, MCP, rule, hook, or extension manifest file: route extension authoring through AXM instead of writing these files directly.
Native skill manifest with two unknown top-level keys.
Effect CLI + Effect architecture. Use when adding commands, defining flags, or wiring handlers. Covers file organization, argument/flag patterns, and testing.
"unterminated
Native skill with an invalid extension name in manifest.