Guide · Agent skill
Use the planning-with-files-zht agent skill (Traditional Chinese)
Updated · written with AI from the project's own documentation and checked against our catalogue data
planning-with-files-zht is an agent skill that gives an AI coding or research agent a persistent working memory made of three Markdown files saved in your project: task_plan.md, findings.md and progress.md. This guide is for readers who have never installed an agent skill and want their agent to plan, track and resume long multi-step tasks in Traditional Chinese. By the end you will know how to install the skill, how to trigger it with everyday requests, what it needs to run, and which safety limits to keep in mind.
Planning with Files (Traditional Chinese)OthmanAdi/planning-with-files · MIT Download the skill Quick start
- 1Download and unzip. Download the ZIP for skills/i18n/planning-with-files-zht from this page and unzip it.
- 2Place the folder. Put the unzipped folder in ~/.claude/skills/ for personal use, or in your project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Ask with a trigger phrase. Describe a multi-step task and use a phrase such as 幫我規劃 or 拆解專案 so the skill is selected.
- 5Check the plan. Let the agent initialise the task folder, then validate task_plan.md with the bundled check script before implementation starts.
- 6Review and resume. Read task_plan.md, progress.md and findings.md to see status and continue later from the same files.
What an agent skill is and what this one does
An agent skill is a folder with a SKILL.md file and supporting files that an AI agent loads when a task matches the skill's description. This one, from the planning-with-files project under the MIT licence, is for complex work that would otherwise be lost in the chat history, such as research or tasks that need more than about five tool calls.
When active, the agent keeps the plan on disk instead of only in its context window. It writes a plan file with phases and status, a findings file for research and decisions, and a progress file for the session log, test results and errors. It re-reads the selected plan before major decisions so the goal stays in view.
- task_plan.md: phases, progress and decisions, updated after each phase
- findings.md: research notes and discoveries, updated after any finding
- progress.md: session log, test results and error log, updated throughout the session
- Lifecycle hooks inject the selected project plan into the agent's context
- Chinese-language templates and trigger phrases are included
Install the skill from the ZIP or the Claude apps
Download the ZIP for skills/i18n/planning-with-files-zht from the page you are reading, then unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder so it applies only there.
If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills, then upload
Trigger the skill with natural requests
You do not need a special command. Ask in ordinary language and include one of the skill's trigger phrases. The description lists these phrases: 任務規劃, 專案計畫, 制定計畫, 分解任務, 多步驟規劃, 進度追蹤, 檔案規劃, 幫我規劃, 拆解專案.
A good request names the task and asks for a plan. For example, you might write 幫我規劃這個專案的研究流程 (help me plan the research workflow for this project) or 請幫我拆解專案成幾個階段 (please break this project into phases).
- 幫我規劃這個專案的研究流程
- 請幫我拆解專案成幾個階段
- 制定計畫來完成這次的程式重構
- 用多步驟規劃追蹤這個任務的進度
What happens when you start a new planned task
For a new task, the skill directs the agent to run its init-session script with a task name, then use the printed PLAN_ID to pin the plan. Phases must keep the format ### 階段 N：名稱 with exactly one status line. The skill also provides a validation check that you can run on a plan file. The check verifies format only, not whether the task is finished or whether a hash attestation is valid.
Existing plans are kept. If the format check fails, the agent should explain the problem to the plan owner rather than rewriting the file automatically.
- Phase status values are pending, in_progress and complete
- If the helper scripts cannot run, the agent should copy the bundled templates unchanged and say that no native format check was performed
sh "$PWF_SKILL_DIR/scripts/check-complete.sh" --validate path/to/task_plan.mdOptional: review past sessions only when you ask
Automatic recovery reads only the project planning files. The session-catchup.py script and the lifecycle hooks do not look at stored agent session history unless you explicitly ask for it. When you do ask, the script can run in metadata mode, which reports whether there is resumable activity for the same project without printing transcript content, tool commands, paths or session IDs.
Replay mode is optional and bounded. Its excerpts are wrapped with a nonce and must be treated as untrusted data, not as instructions.
- Metadata mode reports activity counts only
- Replay mode prints a limited excerpt for the same project and needs your explicit consent
python3 "${SKILL_DIR}/scripts/session-catchup.py" --metadata "$(pwd)"Requirements and safety notes
The helper scripts need POSIX sh on Linux or macOS, or PowerShell on Windows, and Bash for check-complete.sh and related helpers. Python 3 is needed only when session-catchup.py is used. The skill runs scripts and includes lifecycle hooks that run on tool calls. The source states that the skill makes no network uploads and that it never executes commands declared inside Markdown.
Because hooks run on every tool call, review the scripts before installing in a sensitive project. An optional gate mode asks the agent to continue only where the host supports it. If the hook script is not found, plan injection is turned off and a notice is shown once per prompt. Setting PWF_SCRIPT_DIR to the skill's scripts directory, or installing to a user-level path, fixes this.
- Runs scripts: yes
- Uses network: no
- Hooks run on prompts, tool calls, stop and before compaction
- Review the scripts folder first in sensitive projects
Tips for getting the most from the skill
Use a separate plan folder for each parallel task by pinning PLAN_ID before starting each host, or by using separate git worktrees. Exporting a variable inside a subprocess does not change the host environment. Keep one plan owner; workers should report through their own ledger or assigned file rather than rewriting the shared plan.
Skip the skill for short questions, single-file edits and quick lookups, since the planning files add little value there.
- Re-read the plan before important decisions and after an interruption
- Record every error in the plan file so the same failed step is not repeated
- If the user asks for more work after all phases are complete, add new phases and a new progress entry
What people use it for
- Multi-week research where findings must be saved before the context fills up
- A coding refactor split into phases with a log of tests and errors
- Writing a long report with a visible outline and progress record
- Running several parallel tasks, each with its own pinned plan folder
Frequently asked questions
Is the planning-with-files-zht skill free to use?
The source identifies the licence as MIT. Check the repository licence text for the full terms before reusing or redistributing it.
How do I install planning-with-files-zht?
Download the ZIP for skills/i18n/planning-with-files-zht from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload it under Settings, Capabilities, Skills.
What does planning-with-files-zht need to run?
It needs POSIX sh on Linux or macOS, or PowerShell on Windows, and Bash for check-complete.sh and related helpers. Python 3 is only needed when you explicitly use session-catchup.py.
Does the skill send my files to the internet?
The source says the skill has no network upload path and that it uses no network. It does run local scripts and lifecycle hooks, so review them before use in sensitive projects.
Will the skill read my past agent sessions automatically?
No. Automatic recovery reads only the project planning files. Session history is checked only if you explicitly run session-catchup.py, and replay output should be treated as untrusted data.
How is this different from the English planning-with-files skill?
The source describes this as the same workflow with Chinese-language templates and trigger phrases. Choose the Traditional Chinese version if you want Chinese templates and trigger words.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you run long research or coding tasks in Traditional Chinese and want a visible, editable plan you can resume later. Pick something else for short questions, quick edits, or sensitive projects where you do not want hook scripts running on every tool call without review.
Download the skill