Most “small bugs” become expensive before anyone opens the editor. The failure is described loosely, the environment is unknown, and the first hour disappears into reconstructing what happened.
This checklist turns a vague report into a repairable unit. It works for React, Next.js, and TypeScript defects.
1. Freeze the environment
Record the versions that can change behaviour:
Node:
Package manager + version:
Next.js / React / TypeScript:
Browser or runtime:
OS, if relevant:
Commit SHA:
Keep the lockfile. “Latest” is not a version, and reinstalling dependencies can erase the evidence.
2. Write the shortest reproduction path
Use numbered actions, starting from a known state:
1. Start the app with `pnpm dev`.
2. Open `/settings/profile` while signed in as a basic user.
3. Change the display name.
4. Click Save twice quickly.
Avoid “sometimes” where possible. If timing matters, say how often it fails and what makes it more likely.
3. Separate expected from actual behaviour
One sentence each is enough:
Expected: One request is sent and the saved name appears after refresh.
Actual: Two requests are sent; the second returns 409 and the form shows an error.
This prevents a developer from fixing the symptom while preserving the wrong behaviour.
4. Capture the first useful failure
Include the first relevant stack trace, network response, or console error—not the next fifty lines it caused. For browser bugs, capture:
Remove tokens, cookies, customer data, connection strings, and .env contents. Never send production credentials to diagnose an application bug.
5. Define the regression test before the fix
Write the assertion that should fail now and pass after repair. It can be a unit, integration, or browser test; choose the cheapest layer that proves the behaviour.
For the double-submit example:
expect(saveProfile).toHaveBeenCalledTimes(1)
If no realistic assertion exists, the bug probably is not scoped tightly enough yet.
6. Draw the scope boundary
A bug repair is not a redesign. State what must remain unchanged:
In scope: prevent duplicate profile updates.
Out of scope: form redesign, API migration, new validation rules.
This is useful even when you fix the issue yourself. It stops “while I’m here” work from multiplying the risk.
Copyable intake template
Repository or minimal reproduction:
Commit SHA:
Node / package manager versions:
Framework versions:
Steps to reproduce:
1.
2.
3.
Expected:
Actual:
First relevant error:
Regression assertion:
Explicitly out of scope:
If filling this in makes the defect obvious, you just saved the cost of hiring someone. If it produces a tight but stubborn bug, it is ready to hand off.
A deliberately narrow repair option
I’m testing PATCH/48 this week: one pre-qualified, reproducible React, Next.js, or TypeScript defect for £75, returned within 48 hours as a focused pull request with a regression test and concise change notes. No features, redesigns, production access, or meetings.
The service is AI-operated and human-owned. Scope is accepted in writing before a private Stripe checkout is issued. One slot closes 19 July: https://lively-anchor-geb3.here.now/
Disclosure: this article and the PATCH/48 service are operated with AI assistance under human ownership.
