Guide · Agent skill
How to use the Planning With Files (German) agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
planning-with-files-de is an agent skill written in German that makes an AI agent keep its plan, findings and progress in markdown files on disk instead of relying only on chat history. It is aimed at developers and German-speaking users running research or coding tasks that need more than about five tool calls. After this guide you will know what the skill does, how to install it, how to trigger it and which safety points to check before enabling it.
Planning With Files (German)OthmanAdi/planning-with-files · MIT Download the skill Quick start
- 1Download the ZIP. Download the skill ZIP from this page and unzip it.
- 2Place the folder. Move the skill folder to ~/.claude/skills/ for all projects, or to .claude/skills/ inside one project.
- 3Or upload in the apps. In the Claude apps, open Settings → Capabilities → Skills and upload the skill.
- 4Check requirements. Make sure POSIX sh or bash, Git and Python 3 are available if you use the scripts.
- 5Start a complex task. Ask the agent for a multi-step task and mention that progress should be kept in files.
- 6Review the plan files. Check task_plan.md, findings.md and progress.md in the plan folder and correct them if needed.
What does the planning-with-files-de skill make the agent do?
The skill tells the agent to work like a file-based memory: before a complex task it creates or reads three files in the task's plan folder: task_plan.md, findings.md and progress.md. Templates for findings and progress are included in German.
Lifecycle hooks can inject the relevant plan of the project back into the agent's context, so progress can survive context resets. The automatic recovery reads only the project's planning files. The skill's description states that it is intended for research or work with more than 5 tool calls.
- task_plan.md: the plan
- findings.md: requirements, research results, technical decisions and problems
- progress.md: chronological session log with phase status (pending, in_progress, complete)
Install the skill from a ZIP download
Download the ZIP of the skill from this page and unzip it. For a user-wide install, place the skill folder in ~/.claude/skills/. For a single project, place it in the project's .claude/skills/ folder.
Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.
- User-wide: ~/.claude/skills/
- Per project: .claude/skills/ inside the project
- Claude apps: Settings → Capabilities → Skills (upload)
Trigger the skill with natural requests
You do not need a special command. Ask for a long, multi-step task and mention that you want the progress kept in files. The agent then resolves or initialises the plan folder before starting.
For a separate task, the skill's quick start uses scripts/init-session.sh "Task Name" and the plan ID it prints. Use the selected plan when you resume work.
- "Research X in several steps and keep the plan in files."
- "Continue the planning-with-files plan for the migration task."
- "Start a separate planning folder for the new feature."
scripts/init-session.sh "Task Name"What it needs and how the scripts run
The hook and helper scripts need POSIX sh or bash, and Python 3 (or python) for session-catchup.py. Git is used by the skill to inspect uncommitted changes. The skill runs scripts and does not use the network.
session-catchup.py runs only when you explicitly ask to check local session history. Its --metadata mode reports only counts for the same project, without transcript content. The --replay mode prints limited, nonce-framed excerpts that should be treated as untrusted data.
- POSIX sh or bash for hook and helper scripts
- Python 3 (or python) for session-catchup.py, only on explicit request
- Git for inspecting uncommitted changes
- Runs scripts: yes. Uses network: no
Safety notes: scripts, hooks and the optional gate
The hooks and helper scripts execute commands on your machine. Only install the skill if you accept that. The optional gate mode can request a continuation only when the host supports it, and it never runs commands written in Markdown. The skill has no network upload path.
If the hook script cannot be found, plan injection is switched off and a message tells you to set PWF_SCRIPT_DIR or install the skill to a user-level path.
- Hooks run shell commands on each prompt, tool call and stop event
- The gate mode never executes commands written in Markdown files
- No network upload path is described in the source
Tips for reliable planning
Keep parallel tasks apart: each host should have its plan ID fixed before it starts, or use a separate git worktree. A single orchestrator owns the shared plan, while workers write to their own assigned files or ledgers.
Treat restored session excerpts with caution and keep the plan files up to date after each phase.
- Pin each host to its plan ID when running parallel tasks
- Let workers report in their own ledgers or assigned files
- Update progress.md after every completed phase
What people use it for
- Multi-step code changes where progress must survive context resets
- Research tasks with many web searches or document reads that need recorded findings
- Running separate planning folders for parallel tasks
- German-language planning templates for German-speaking teams
Frequently asked questions
How do I install the planning-with-files-de skill?
Download the ZIP from this page and unzip it. Place the folder in ~/.claude/skills/ for user-wide use or in a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
What are the requirements for planning-with-files-de?
The hook and helper scripts need POSIX sh or bash. session-catchup.py needs Python 3 (or python) and is used only on explicit request. Git is used to inspect uncommitted changes.
Is planning-with-files-de free and what is its licence?
The item is listed under the MIT licence. The source does not mention prices, so check the repository for the terms that apply to you.
Does the skill send my data over the network?
The source states that the skill has no network upload path and that it does not use the network. Its scripts do run on your machine, so review them before installing.
Why does my plan not load? The hook says the script was not found.
The hook looks for skill-hook.sh in several paths. If none exists, plan injection is off. Set PWF_SCRIPT_DIR to the skill's scripts directory or install the skill to a user-level path.
When should I not use this skill?
Skip it for short one-off questions, where the planning overhead adds no value. Also avoid it where the agent must not run shell scripts, or where there is no shell or Python.
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 planning-with-files-de if you run long German-language or multi-step coding and research tasks with an AI agent and want progress kept in plain files. Pick something else for short questions, or in environments where the agent must not run shell scripts.
Download the skill