菜单
Creates and improves Claude subagents (agents/<name>.md) to the Advanced Skill Library Standard. Use when you need to author a new subagent, scaffold an agents/ delegate, declare its tools and chain, or raise an existing subagent's conformance.
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/product-on-purpose/agent-skills-toolkit --skill askit-build-subagent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。
Creates and improves a plugin's documentation across modes (readme, quickstart, tutorial, how-to, reference, glossary, faq, troubleshooting, architecture, folder-readme, and an Astro Starlight docs site) to the Advanced Skill Library Standard. Use when authoring or refreshing docs, scaffolding a folder README, standing up a docs site, or aligning documentation with the component index.
Surveys what the agent platforms shipped since this repository last looked, diffing each vendor's release feed forward from a recorded version pin, and writes a dated survey record without deciding anything. Use when asking what is new in Claude Code, Claude Cowork, Codex or the agentskills.io spec, before planning a release that claims to be current with the vendors, or on a periodic currency sweep.
Assesses a new agent capability against the capability matrix, the Standard, and every skill, subagent and command this plugin ships, then proposes a matrix update, a backlog entry or an ADR draft without implementing any of them. Use when deciding whether a new Claude Code, Cowork or Codex capability should be modelled or adopted, when analysing a capability gap after a survey reports one, or when checking which components were last verified against an older agent version.
基于 SOC 职业分类
正在显示 SKILL.md
| name | askit-build-subagent |
| description | Creates and improves Claude subagents (agents/<name>.md) to the Advanced Skill Library Standard. Use when you need to author a new subagent, scaffold an agents/ delegate, declare its tools and chain, or raise an existing subagent's conformance. |
| metadata | {"version":"0.2.0","tier":"universal","audience":"intermediate"} |
Author conformant subagents. Two modes: create scaffolds a new agents/<name>.md (frontmatter + role body); improve raises an existing subagent toward the quality bar. Subagents are Claude-only for plugin distribution (Codex v0.135 plugins cannot ship [agents.*]; Standard sec 3.3), so this skill emits no Codex artifact. Authoring depth is in references/authoring-subagents.md.
When the user asks to create, scaffold, write, or improve a subagent (an agents/<name>.md delegate).
templates/agent.md into agents/<name>.md.name equal to the file basename; a description that states what AND when (Standard sec 8.1); tools as the narrowest set (sec 9); optional model; metadata.chain naming the components it may invoke as a comma-separated string (the recommended shape - a YAML list is silently coerced to a string repr by the reference implementation, see authoring-chain-contracts.md; omit chain if there is none); metadata.version, metadata.tier, metadata.status, and metadata.agent-targets: [claude] (sec 3.3 - subagents are Claude-only for plugin distribution).hooks, mcpServers, or permissionMode. Claude Code states plainly that "for security reasons, hooks, mcpServers, and permissionMode are not supported for plugin-shipped agents" (plugins reference, read 2026-08-12). Writing one does not configure anything; the runtime refuses it, and the author is left believing otherwise. The supported set is name, description, model, effort, maxTurns, tools, disallowedTools, skills, memory, background, and isolation (whose only valid value is worktree). Marketplace scope reports this across every member of a catalogue.library.json components.subagents as { name, path, version, tier, status }.chain, add the entry to agents/_chain-permitted.yaml (<name>: [<invoked>, ...]) so S4 has no orphan.node scripts/evaluate.mjs . --json and iterate until 0 errors (S3 components index + S4 chain contract must be clean).node scripts/evaluate.mjs . --json and read the report.components.subagents to match disk; an S4 orphan -> add the missing caller -> callee to agents/_chain-permitted.yaml; an S4 phantom (the contract names a component not on disk) -> remove that contract entry, fix a misspelled name, or create the missing component. For any other finding, read its message and apply the fix it states.Claude-only: a subagent is agents/<name>.md (Claude auto-discovers it). Codex does not ingest plugin-shipped subagents (Standard sec 3.3), so there is no Codex render. Authoring is performed inline by the running agent; the toolkit's own askit-skill-author subagent is the delegated authoring role for skills.