菜单
Create, audit, or package a portable Claude/Codex skill. Use when building a new skill, converting an existing workflow into a reusable skill, reviewing whether something is a real skill or just a rule/reference/kit, or preparing a skills repository catalog.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/ariaxhan/the-agent-library --skill create-portable-skill命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Turn a borrowed metaphor, analogy, or cross-domain inspiration into a testable mechanism, or prove it is decoration and discard it. Use when a design leans on language from another field ("immune system", "evolution", "market", "quorum", "crystallization"), when someone proposes copying how another domain solves a problem, or when reviewing a system whose names sound deeper than its code. Produces either a named domain-free invariant with evidence and falsifiers, or an explicit rejection with the reason it failed.
Design systems so a failure can never present as success. Use when building or reviewing pipelines, extractors, reconciliations, checkers, or any computation whose output humans act on; when missing inputs, unreadable files, or crashes could pass silently; when a tool "found" an explanation that feels too convenient; or on "make this fail loudly", "could this lie to us", "harden the failure paths". Produces named invariants, explicit skip and no-answer states, honest cause classes, and adversarial regression fixtures.
A compact decision protocol for non-trivial choices, reviews, governance calls, and live disagreement. Use implicitly at decision points, even when nobody says "debate", whenever a choice is high-impact, hard to reverse, policy-bearing, or has multiple plausible paths that evidence alone cannot settle. Skip for trivial, cheap-to-reverse, or deterministically decidable choices.
正在显示 SKILL.md
基于 SOC 职业分类
| name | create-portable-skill |
| description | Create, audit, or package a portable Claude/Codex skill. Use when building a new skill, converting an existing workflow into a reusable skill, reviewing whether something is a real skill or just a rule/reference/kit, or preparing a skills repository catalog. |
Create skills that another agent can actually use. A skill is a packaged workflow, not a rule, essay, prompt dump, or product pitch.
Before writing files, decide what the artifact is:
| Type | Definition | Destination |
|---|---|---|
| Packaged skill | Multi-step workflow, concrete output, portable dependencies | skills/<name>/ |
| Draft skill | Real workflow but incomplete body, name, trigger, or bundled files | skills/<name>/ with explicit draft status |
| Kit | Installable system made of hooks, templates, multiple files, or multiple skills | kits/ |
| Rule/pattern | Principle, invariant, hook pattern, or one-step guardrail | rules/ or docs, not skills/ |
| Reference | Background methodology a skill can point to | reference/ |
| Reject | Too vague, too local, not a workflow, no crisp trigger | Do not catalog |
If it could be one bullet in a rules file, do not make it a skill.
A packaged skill must pass all five:
description: names real user phrasings and when to use it.Use lowercase hyphen-case. Prefer verb-led names that say the output:
write-project-vision-handoff, critique-vision-handoff, hunt-research-sourcesFolder name and frontmatter name: must match exactly.
The frontmatter description: is the discovery surface. It must include:
Use third person and be specific. The body loads only after the skill triggers, so do not hide trigger logic in the body.
Minimum:
skills/<name>/
SKILL.md
agents/openai.yaml
Add bundled files only when they directly support execution:
reference/ for templates, output formats, checklists, or long examplesscripts/ for deterministic repeatable operationsassets/ for templates or media used in outputsDo not add README, changelog, installation guide, or extra docs inside the skill folder.
Check:
SKILL.md starts with YAML frontmatter.name: and description:.name: matches folder.agents/openai.yaml has a short display name, short description, default prompt, and invocation policy.Add the skill to exactly one primary category. If it is draft, local-wrapper, kit, rule, reference, or reject, say that honestly. The catalog is a shelf, not a dream board.