npx maisongr-skills
Why this exists
AI coding agents rarely fail for lack of knowledge. They fail because finishing the task pulls harder than doing it right. So they:
- guess what you meant instead of asking;
- build abstractions and options the task never needed;
- "improve" files nobody mentioned;
- report "all tests pass" without running them;
- leave dev servers running in the background, where you never see them, until the laptop freezes;
- push, drop a column or break a public API when nobody asked them to.
And when your team uses three different agents, you get three different standards, or none.
Maison GR gives every agent the same seven principles and thirteen task methods, from one source file, installed with one command.
60 seconds
- In any project, run
npx maisongr-skills.
- Ask your agent for something slightly ambiguous: "Add caching to the users endpoint."
- Instead of guessing, it asks what to cache and for how long, or states the assumptions it made.
- When it finishes, it reports what it did and what it could not check. For example:
Changed: src/api/users.ts: added a 60 s cache to GET /users
Verified: npm test (42 passed), npm run lint (clean)
Not verified: cache invalidation under concurrent writes (no integration environment)
Decisions: chose in-memory over Redis; revisit if you run more than one instance
Undo it any time with npx maisongr-skills remove.
Receipts: "tests pass" you can check
An agent can write "Verified: npm test (42 passed)" without having run anything. With the Claude Code plugin installed, Maison GR checks.
- Every shell command the agent runs in a session is recorded, with its exit code.
- When the agent tries to finish, each command listed under Verified is checked against that record.
- If a command never ran, or its last run failed while the report says it passed, the agent is sent back to run it and report the real result, or to move it to Not verified.
- It also catches guarded, irreversible actions the report doesn't own — a force-push,
git add -A, a history rewrite, a --no-verify, or a Git config change — and sends the agent back to disclose and justify them, or undo them.
- It catches servers left running: a
next dev, vite, wrangler dev or npm start the session started and never stopped, and it stops them itself when the session ends.
Maison GR receipts: your report lists commands under "Verified" that this session's record does not support:
- `npm test`: no command like this ran in this session.
Run them now and report the real result, or move them to "Not verified" with the reason.
The same record catches guarded actions taken silently (see principle 7, Guard irreversible actions). If the session force-pushed, skipped Git hooks with --no-verify, staged everything with git add -A, rewrote history (amend, rebase, reset --hard), or changed Git config, and the report never mentions it, the agent is sent back to own and justify the action — or undo it.
Maison GR receipts: this session took a guarded action without owning it in the report:
- this session force-pushed (`git push --force origin fix/foo`) but your report never mentions it.
Each is allowed only if the user asked for it here. Say so and justify it in your report, or undo it.
And it catches servers left behind (see principle 5, Verify before completion). A dev server, watcher or tail has no window, so a forgotten one keeps running after the session, and a few of them freeze a laptop that is short of memory.
- It knows which servers are the agent's. Just before a server command runs, Receipts tries the ports it may use: its own (
--port, -p, PORT= or the tool's default, such as 3000 for Next.js or 8787 for Wrangler) and the next ten, where Vite and Next.js move when it is taken. A port that already answered belongs to someone else, so a server you started yourself is never blamed or stopped. A server whose port is unknown (npm run dev, wrangler tail) is checked against Claude Code's own list of background tasks still running.
- Before the agent finishes, anything of its own still running sends it back to stop it, or to list it under Left running with its address when you asked to keep it. Writing "I stopped it" is not enough while the port still answers.
- When the session ends, Receipts stops the agent's servers that are still running, except those it left running on purpose.
- When the next session starts, it stops what a session that ended without that cleanup (a crash, a closed window) left behind, and tells the agent, so it can tell you. It stops only the very process it recorded, matched by process ID and start time: a program that later reused the port is never touched.
Maison GR receipts: this session started a server that may still be running:
- this session started `npx next start -p 3100` and it is still listening on localhost:3100.
Stop it now. If the user asked to keep it, list it under "Left running" with its address and how to stop it.
It checks only command names in backticks (and the guarded Git actions and servers above), ignores what is inside a quoted commit message, never blocks twice in a row, doesn't raise an action again once a report has owned it, keeps its record on your machine, in Claude Code's plugin data folder, and connects only to localhost. To stop a server it looks up the process on that port with the system's own tools (PowerShell on Windows, lsof or ss and ps on macOS and Linux); if it can't find or identify the process, it reports the server and stops nothing. Receipts need the Claude Code plugin; other agents get the same report format without the check.
여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗
