QIMING // PROJECT SYSTEM001 READY
PROJECT-LOCAL / USER-OWNED

Give every directory
a way of working
of its own.

New to AI? You do not need a perfect prompt. Put your material in one folder and tell an agent the first thing you want to do. Qiming helps it understand the workspace and leave a clear place to continue.

See installationExplore an example

One directory for one continuing body of work. Existing files stay where they are.

QIMING / WORKSPACE CONTROLLOCAL INSTANCE
01 / INPUT

Current directory

  • Existing files
  • Work in progress
  • Your conventions
Understand · Connect · Evolve
02 / OUTPUT

Your Qiming instance

Work entryMembersKnowledgeRecovery
ONE PROJECT / ONE OWNER / MANY AGENTSQ/01

Change agents or computers without explaining your work from scratch.

01 / OPERATING MODEL

Adopt once. Grow with your actual work.

Look at what you have, then do one concrete thing. Qiming keeps the useful result in your directory without asking you to design a taxonomy first.

01

Read the room

Inspect the directory, project instructions, existing material, and current goal. Your names and structure come first.

02

Create an entry

Establish a project-owned instance, short handoff, and agent instructions so a new session knows where to continue.

03

Keep the outcome

Retain tasks, errors, knowledge, scripts, Skills, and MCPs according to their real maintenance value.

MEMBERS / WHAT STAYS

A “member” is a managed object, not a subscription.

Projects, accounts, devices, reusable scripts, Skills, programs, and MCPs can each have an identity, owner, and handoff record when you maintain them over time. Internal files stay with their parent. Passwords and keys are represented only by references to secure storage.

Stable identityOrigin and ownerReal verificationNext step
FOR FIRST-TIME AI USERS

Your first AI project starts here.

Prepare a folder and put your material in it. After installing the Qiming seed, explicitly ask your agent to adopt it in that folder. Then describe one task in plain words.

01

Prepare a folder

An empty folder works for a new project. Keep an existing project where it is.

02

Introduce it to the agent

Say “Enable Qiming in this directory.” The agent reads the material and creates a project entry.

03

Do one small task

Organize a chapter or fix a problem. Leave the outcome, verification, and next step.

Try this sentence

This is my first time using AI to manage this folder. Tell me what is here and suggest one small first task. Wait for my confirmation before changing anything, then leave the result and next step here.

02 / REAL WORK

Two real folders. Six ways to make it yours.

The first two cases come from my real work, with project, account, and tool names removed. Both my computer management directory and the health project root have adopted Qiming. The old health development directory is an archive; starting there does not automatically load the main project. Six more cards show transferable uses.

Computer careQiming adopted

A development computer rebuilt again and again

The problem

I often reinstall my operating system. Software, scripts, and account information were scattered across conversations and folders, so every restart needed another explanation.

What happened

I gave the computer its own management directory. Qiming checked the current system and existing records, registered a long-lived media tool as a member, and tracked installation separately from functional verification.

What remains

A short entry now points to the system check, tool record, installation result, and next task. Account records contain secure-storage references, not passwords.

Verified boundary

The service and management page were connected. Actual generation is still unverified, and backup recovery remains to be checked.

Ask your agent

$qiming Continue managing this computer. Check the current system and existing records, decide whether this new tool needs a member record, and separate installation, real function testing, and next steps.

Complex developmentQiming adopted in the main project

A health project with a contract and several roles

The problem

This health project combines contract scope, two active codebases, several roles, and a customer preview. An old development directory remains for history and can be mistaken for the live delivery location.

What happened

I adopted Qiming at the main project root. It preserved the existing rules and connected both codebases, delivery evidence, and the current task through a short entry. A project Skill and AGENTS.md give new sessions their starting point.

What remains

The main directory now has its own instance, work records, and asset ownership. Verified preview work is tracked separately from remaining contract requirements. The old directory is an archive and is not used for releases.

Verified boundary

The main instance and Codex binding passed checks. The old directory has no separate Qiming instance. A customer preview is not final contract acceptance, and sessions started inside either codebase still need a scope check.

Ask your agent

$qiming Continue from the main project root. Read its entry, existing rules, and current task; check both active codebases and delivery evidence. Do not develop from the archive, and track the preview separately from final acceptance.

01Writing

Turn scattered material into chapter one

Say this

This folder has old drafts, interview notes, and ideas. Inventory the material and sources, agree on the question for this chapter with me, draft chapter one, and leave revision notes and the next chapter task.

Qiming keeps track of

A draft, source trail, open questions, and a next-chapter entry.

02Site operations

Plan a week of content updates

Say this

Review the site categories, recent posts, and publishing flow. Plan three posts with goals, owners, and dates. After the first, record the preview, live result, and follow-up checks.

Qiming keeps track of

Site status, content schedule, publication evidence, and review task.

03Learning

Organize a course from zero

Say this

Inventory these notes and exercises, identify what I know, and propose a seven-day plan. Do only lesson one today; record missed questions and tomorrow’s review.

Qiming keeps track of

Learning goal, daily steps, mistakes, and review trail.

04Open source

Revive a repository left idle

Say this

Read the README, open Issues, tests, and recent changes. Pick one bounded problem, fix it, run the relevant checks, and list older problems still unverified.

Qiming keeps track of

Current state, one verified change, and remaining issues.

05Small team

Turn a client request into a deliverable

Say this

Read the request, meeting notes, and delivered files. Separate agreed items from open questions, then list the owner, acceptance condition, and missing client input for the next delivery.

Qiming keeps track of

Source of requests, ownership, acceptance, and blockers.

06Research

Turn research files into an answer

Say this

Organize the papers, links, and experiment notes by question. Cite the source and uncertainty for each conclusion, then propose the next test.

Qiming keeps track of

Traceable conclusions, evidence, and untested hypotheses.

03 / PROMPT LIBRARY

Useful prompts

Use these in the current project after adoption. Its Qiming instance belongs to this project only.

01

First adoption

$qiming Enable Qiming in this directory. Read existing files and project instructions, preserve the structure, and create a handoff for this project. Tell me the goal and next step you found.

02

Continue previous work

Continue the unfinished work in this project. Read its entry and relevant records, advance the current goal, and leave verification results and the next step.

03

Retain a lesson

Review the error and resolution from this task. Decide what belongs in reusable knowledge or a tool, and record its source, scope, and verification.

04

Manage an asset

Add the account, device, or script I explicitly handed over to this project. Record its purpose, owner, and maintenance entry. Keep credentials in secure storage; record only a reference.

QIMING / CLEAR ANSWERS

Frequently asked questions

Understand what Qiming does and where its project boundary sits before you begin.

What is Qiming Skill?

Qiming is a project-local AI work management Skill. It reads the files and rules in a directory you choose, then creates an entry so future agents can continue development, writing, site operations, or other long-running work.

How do I start if I am new to AI?

Prepare one folder and put your material there. Install the Qiming seed, then ask your agent in that folder: “$qiming Enable Qiming in this directory.” Give it one small task to do first.

Does installation automatically manage all my folders?

No. Installing the seed only makes adoption available. An independent instance is created only in a project you explicitly select. Existing names, files, and rules take priority.

Does “member” mean a paid subscription?

No. A member is a project, account, device, reusable script, Skill, program, or MCP that you maintain over time. Retained work records its owner and source; internal files stay with their parent member.

Where do my data and passwords live?

Project rules, tasks, knowledge, and tools remain in your own project directory. Account records refer to secure credential storage; ordinary project files do not contain passwords or keys.

Which agents have been verified?

Project adoption, binding, and remote installation have targeted verification with Codex. Claude Code, Gemini CLI, Cursor, OpenCode, and other operating system combinations still require individual real-world tests.

04 / ORIGIN

The Qiming story

My agent Dark源 and I first wanted a better way to manage our own projects. In real work we gradually kept project personas, directories, rules, scripts, and lessons. They became a management template. Qiming grew from that template.

I switch tasks and agents, and I often reinstall my operating system. The hard part was not reinstalling software; it was losing context that lived only in the previous conversation. Dark源 and I moved the entry, decisions, and next step back into the working directory.

I named it Qiming for the morning star: a point of direction before sunrise. I want someone using AI for the first time to start with a single folder and one small task, without learning a vocabulary of tools first.

The install is only a seed. The members, knowledge, tools, and working habits that follow belong to the person who uses it.

05 / CHANGELOG

Changelog

Official domain and search access

The official site moved to qiming.dashen.wang with direct Cloudflare Pages hosting and proxying enabled. All four language pages include search metadata, a sitemap, and visible FAQs.

Project scope and continuity

Improved project-owned instances, startup context, agent instructions, and ownership of derived capabilities. Evidence records 112 automated checks and a targeted Codex trial; the full cross-agent matrix remains untested.

A usable management seed

Established the Skill entry, deterministic tools, source mapping, member and knowledge records, export, and recovery rehearsal.

06 / GET QIMING

Begin in the directory you're in.

Install the Qiming seed into your AI agent, then explicitly adopt it in the target project. First use requires Python 3.11+. Installing the seed does not change your other projects.

QIMING / INSTALL TERMINAL
Install the seed with one command
npx skills add cat9999aaa/qiming-skill --skill qiming
Copy this installation prompt to your agent

Install the qiming Skill from the official GitHub repository https://github.com/cat9999aaa/qiming-skill into the current project. Check Python 3.11+ first. Preserve existing files and project instructions. Do not adopt the project until I explicitly ask you to enable Qiming here.

View official source
After installation, tell your agent in the target directory:

$qiming Enable Qiming in this directory. Preserve the existing material and structure, then create an independent instance and the next step for this project.

The command uses project installation through the Skills CLI. Actual loading depends on the agent. Claude Code, Gemini CLI, Cursor, OpenCode, and other combinations still need individual real-world verification.