菜单
Find and implement the next incomplete task from .llm/todo.md.
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
Restore claude/codex agents from a tmux-reboot snapshot after tmux-resurrect reopens their windows.
Snapshot running tmux claude/codex agents immediately before a reboot so their sessions can be restored afterward.
Commit changes, run git-test precommit checks, and fix failures. Use when asked to fix build, test, lint, typecheck, or precommit failures.
Running precommit checks and build validation. ALWAYS use after ANY code changes.
This skill should be used after completing any task, before returning control to the user. Always run this skill — it handles the case where there's nothing to do.
Coordination guidelines for leading a team of task-executing agents
| name | markdown-do-one-task |
| description | Find and implement the next incomplete task from .llm/todo.md. |
Use the markdown-tasks skill for script path rules and task semantics.
Extract the next task:
python <plugin-root>/scripts/task_get.py .llm/todo.md
If no task is returned, report that there are no incomplete tasks.
Implement only the extracted task. Ignore unrelated TODO/TASK comments in source code.
Run appropriate tests and validation. Complete the repository's finish workflow so the task has its own commit, precommit passes against that commit, and the branch is rebased when required. If the build, git, and orchestration plugins are installed, use their precommit, git-commit, git-rebase, and finish skills.
Mark the task complete only after implementation, commit, and validation succeed:
python <plugin-root>/scripts/task_mark.py .llm/todo.md
If implementation or validation fails, leave the task incomplete and report the failure. The markdown-do-all-tasks leader is responsible for marking failed attempts [!].