Project Skeleton

project-skeleton

Use this as a short entry point for research-project-os.

If .project_os/ exists in the current project or a parent directory, never create task_plan.md / findings.md / progress.md as the primary memory and never start a parallel project-state workflow. Resume the harness first.

Route

1. Detect the project root.
2. For a machine-readable plan, use project_os.py route --root <project> "项目骨架" or project_os.py route --root <project> "开工" first.
3. If .project_os/ is missing, use the research-project-os skill and run its project_os.py new-project flow as a dry-run first unless the user explicitly asked to apply changes.
4. If .project_os/ exists, use the research-project-os skill and run its project_os.py start flow to load project state, optional current_session, active branch/task/run, and context manifest.
5. Keep this skill thin: do not duplicate the harness contract; defer to research-project-os references and scripts.

Preferred user-facing phrases

Map these to the flow above:

• 新项目骨架 / 搭项目骨架 / 初始化项目骨架 -> bootstrap a new .project_os/ harness.
• 项目骨架 / 项目工作流骨架 -> auto-detect bootstrap vs resume.
• 开工 / 继续项目 -> resume existing .project_os/ state.
• 大项目 / 逐步推进 / 制定计划 / 拆解任务 / task_plan.md / findings.md / progress.md -> use .project_os branch/task/session state instead of the old planning-file kernel.
• 项目状态 / 写一个项目状态文档 / 更新项目状态文档 -> use .project_os status/summarize-state instead of a parallel project-state workflow.

Router note

This is only the bootstrap/resume alias. Other short phrases such as 新建分支, 新建会话, 切会话, 暂停会话, 恢复会话, 会话清理, 开始运行, 记录结果, and 设为当前结果 should route through research-project-os and references/short_trigger_router.md.