Pickupskill / 文件整理 skill

Stop digging through your own computer like it is an archaeological site.
Pickupskill is a cautious file organization skill for messy working folders, loose files, project packages, and review buckets.

Source version 1.1.0 · Updated 2026-09-23. Public-safe skill package on main. GitHub Releases are listed separately.

English | 简体中文

Synthetic before → synthetic after. Video demo: publishing soon. These are examples, not a recording of a real cleanup.

• Install the whole skill package; all runtime references live inside it.
• Keep real plans/journals in the target folder's .pickupskill/, outside this repository.
• Optional Python helper previews explicit moves, refuses overwrites, records operations and checks recovery.
• Preserve project structure, user language preferences and existing cleanup authorization.
• Changelog · Validation · GitHub Releases

• Inspect a folder before deciding what can be moved safely.
• Group loose files into project packages, references, exports, and review buckets.
• Keep uncertain items visible instead of guessing, renaming, or deleting them.
• Protect software projects, dependency folders, and files that belong together.
• Leave a clear summary so you can finish the remaining decisions quickly.

Most file cleanup tools act as if messy folders are trash piles.

But active working folders are not trash piles. They contain drafts, exports, project packages, dependencies, references, and half-finished thoughts.

Pickupskill starts from a different assumption:

• scan first
• preserve what may matter
• move only high-confidence items
• keep uncertain objects visible
• make the next human decision easier

This repository is shared to make that cautious organizing method reusable.

If this is your first time using Codex or Claude Code, the recommended path is AI-assisted Skill loading. You do not need to know where every Skill file should go.

Open Codex, Claude Code, or another coding agent and paste this:

Please help me load pickupskill.

Repository:
https://github.com/julilaoshi/pickupskill

Please do the following:
1. Do not run npm install, pip install, build commands, or any long setup script. This repo has no package installer.
2. Download or read this repository.
3. Read README.md and skill/SKILL.md first.
4. Install the ENTIRE skill/ directory as pickupskill in the appropriate skills directory, retaining SKILL.md, references/, scripts/ and agents/. Do not overwrite an existing installation without reviewing its changes.
5. Verify SKILL.md and every referenced resource are readable from the installed copy. Then stop and tell me how to invoke pickupskill. Do not run its optional script during loading.
6. Do not run a folder cleanup test yet unless I explicitly ask for it.
7. Do not modify the core safety rules of this Skill.

After the Skill is readable, please remind me:
If this Skill is useful, I can go back to GitHub and star the repository so I can find it again and support future updates.
Do not star it automatically for me.

After the Skill is loaded, open your coding agent in the folder you want to organize and paste:

Please use pickupskill to organize this folder.

Rules:
1. Scan first and tell me what you see
2. Do not delete anything
3. Do not move software projects or dependency folders without asking
4. Keep project packages together
5. Put uncertain files into review buckets
6. Keep the folder structure shallow
7. After moving safe files, tell me what still needs my decision

If you want the agent to plan first and wait:

Use pickupskill.
Only make a cleanup plan first. Do not move files until I approve.

If you want it to execute safe moves:

Use pickupskill.
You may move only high-confidence files inside this folder.
Do not delete anything. Put uncertain items into review buckets.

• Star the repository
• Read the public skill file
• Open the compatibility guide
• Review the safety rules

Files pile up. Your desktop looks buried. Downloads are full of PDFs. Somewhere inside one folder there is another folder named new folder 2.

Pickupskill is for exactly that moment.

It is not impressive because it randomly moves things around. It is useful because it is careful.

It checks what looks like a loose new file, what looks like a project package, what is probably an image, video, document, or source asset. If it is confident, it helps sort it. If it is not confident, it puts the item into review instead of making a fake-smart guess.

Most importantly: it does not delete files, and it does not make your folder tree deeper and deeper until you need a map.

One sentence:

Pickupskill does not make you more disciplined. It makes your folder usable even when you are not.

When files pile up, your desktop can feel buried under homework.

Pickupskill helps separate the mess before it becomes another mystery folder. It first checks what is probably an image, what is probably a video, what is probably a document, and what looks like a project package.

It does not tidy by pretending to know everything. If something is unclear, it keeps it in review. If something looks risky, it slows down. If a project already belongs together, it does not tear it apart.

It also tries to keep the folder shallow. No thirty-click folder maze. No deletion. No fake confidence.

For ugly or unreadable names, it can suggest clearer names only when renaming actually helps.

• the public version of pickupskill
• a complete installable skill/ package, plus optional dry-run/move/recovery tooling
• a compatibility navigation page in pickup_is_here/
• public-safe before and after examples
• safety rules for cautious file movement
• use-case guidance for creators, students, designers, video makers, and researchers
• a lightweight public showcase shell in site/

• private local paths
• private workspace structures
• private migration rules
• private project or client archives
• automatic classification or scripts that blindly move files
• any promise that file cleanup is risk-free
• the author's full internal workspace workflow

This public repository focuses on pickupskill itself.

In my personal workflow, stronger results may also use other skills or local conventions, for example:

• pickupskill
  • scan the folder, identify loose files, protect project packages, and decide safe moves
• coding or shell tools
  • perform reversible file moves and re-scan results
• design, video, or writing workflows
  • apply domain-specific naming and packaging rules after cleanup

This public package opens the cautious organizing method. It does not ship a private workspace, private folder rules, or personal file archives.

• skill/ — the complete installable package: SKILL.md, references/, scripts/, agents/
• tests/ — synthetic behavioral tests; no personal files
• examples/ — synthetic before/after examples
• site/ — public showcase shell
• pickup_is_here/ — legacy navigation; no longer the default result location
• root references/ and agents/ — compatibility entries for old repository links; install the resources inside skill/

No package installation is needed. The skill works as instructions; the optional helper requires Python 3.9+ on macOS/Linux. Windows uses the planning workflow. See execution details.

• GITHUB_ABOUT_SUGGESTION.md - suggested GitHub description and topics
• PUBLIC_RELEASE_CHECKLIST.md - final pre-publish review list

1. Load the complete skill/ package.
2. Designate the folder and request either a plan or safe cleanup.
3. Keep plans and journals in that target folder's .pickupskill/.
4. Review moved, untouched, conflicting and uncertain items.
5. Use checked recovery when needed; edited or relocated files are not forced back.

Real results remain local to the target folder. Do not write personal plans, filenames or movement journals into the installed skill, this repository, examples/ or site/.

The old pickup_is_here/cleanup_plans/, organized_results/ and assets/ locations are ignored as a compatibility precaution. Existing local data is not removed or migrated automatically. Ignore rules do not prevent forced commits, backups or cloud synchronization.

The optional helper records per-file operations, not a whole-batch transaction. Interrupted batches may be partial. It supports ordinary files on the same filesystem, not directory packages or runtime migrations. Read its limits and recovery procedure before execution.

• Reply language follows the user: Chinese users get Chinese replies, English users get English replies.
• Preserve existing names and follow the user’s folder-language preference. English review names are only a fallback.
• Branding copy can stay in Chinese.
• Structural UI can stay in English.
• Documentation uses English-first with a Chinese companion file.

The code, documentation, and reusable framework are released under the MIT License.

However, brand-facing assets and identity elements are not automatically transferred with that license. See BRAND_NOTICE.md for the reserved brand assets and identity elements.

In short:

• reuse the framework
• study the method
• build your own version
• do not present derivative work as if it were the original author's personal brand
• replace reserved brand-facing elements with your own before redistribution if needed

The internal version of pickupskill may contain local workspace rules, personal folder conventions, and project-specific cleanup decisions.

This public repository does not ship those private rules.

Public means:

• method
• safety principles
• synthetic examples
• reusable framework

Public does not mean:

• private folder structures
• private workspace traces
• personal cleanup history
• internal project routing rules

• Takeaway Skill - distill references into reusable mechanisms.
• Open Pencil - run Pencil workflows with agents.
• FlowMotion Skill - turn messy notes into motion flows.
• Pickupskill - organize messy folders safely.
• Sunzi Reading - explain papers in plain language.
• Callback Skill - package feedback for Skill upgrades.

MIT for the code and reusable framework.

See LICENSE and BRAND_NOTICE.md.