菜单
Use subagents liberally for research, exploration, and parallel analysis. Offload work to keep main context clean. Use when context is crowded, tasks can run in parallel, or for research/exploration work. One focused tack per subagent.
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/unional/skills --skill subagent-strategy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。
Bring one of unional's repos or organizations up to the standard baseline — branch protection ruleset, merge and Actions settings, security toggles, and the CI/release file layout. Use when setting up a new repo, when asked to 'update branch protection', 'require all-checks', 'apply my repo settings', 'set up this repo like the others', or to audit which repos have drifted from the baseline.
Apply a consistent documentation standard: controlled plain prose, one home per fact, and every external claim backed by a source. Use when writing or revising documentation, READMEs, or skill bodies; when deciding where a document belongs; when moving, renaming, or splitting a page; when a draft reads as generated and needs its rhythm, filler words, or em dashes fixed; or on requests like 'improve the docs', 'audit the docs', 'this page is too long', 'this sounds like AI', 'humanize this', 'remove the em dashes', or 'where should this be documented'.
Sweep an owner's repos to find which ones stopped publishing, classify each failure by root cause, and report the evidence needed to fix them. Use when asked 'which of my repos are broken?', 'is anything not publishing?', 'audit my releases', 'find repos that need modernizing', or before running modernize-repo so you know which repos to run it on.
基于 SOC 职业分类
正在显示 SKILL.md
| name | subagent-strategy |
| description | Use subagents liberally for research, exploration, and parallel analysis. Offload work to keep main context clean. Use when context is crowded, tasks can run in parallel, or for research/exploration work. One focused tack per subagent. |
| metadata | {"internal":true} |
Use subagents liberally to keep the main context window clean and throw more compute at complex problems.
When the main context is getting crowded or tasks can run in parallel, delegate to subagents.
| Situation | Action |
|---|---|
| Context crowded | Offload to subagent |
| Parallel tasks | Launch multiple subagents |
| Research needed | Delegate to exploration subagent |
| Complex analysis | Split into focused subagent tasks |
