菜单
Use when executing an approved implementation plan and review strictness should adapt to task risk.
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
Use when an approved design needs an implementation plan, and plan depth should adapt to risk and scope.
Safe git operations. Avoids editor-opening commands that hang the agent, blocks destructive actions without confirmation, and enforces safe push semantics. Use when running any git command — commit, tag, merge, rebase, push, reset, or branch manipulation.
Apply portable Elixir architecture and style guidance for maintainable, domain-oriented applications. Use when writing, reviewing, refactoring, or planning Elixir code, especially around module structure, contexts, actions, schemas, workers, queries, file size, and larger codebase organization.
Apply portable React and TypeScript architecture and style guidance for maintainable, feature-oriented frontends. Use when writing, reviewing, refactoring, or planning React code, especially around component boundaries, hooks, state, data fetching, file size, accessibility, tests, and larger frontend organization.
| name | subagents |
| description | Use when executing an approved implementation plan and review strictness should adapt to task risk. |
Execute approved plan tasks via subagents with adaptive strictness. Small or mechanical tasks use a lean medium path; risky or integration tasks escalate to a strict multi-reviewer path with fix loops.
Use the strict path when any of the following are true:
When none of the above apply, default to the medium path.
After each task, report:
At the end of the full plan, produce a final report containing:
