Fix it in minutes, then make sure it never happens again.
TechFix is a pack of small skills for your AI coding agent, one for each thing that goes wrong once AI-built code goes live.
It breaks, it leaks a key, or it runs up a bill. And the agent that built it often said "done" first. Each one has an error people search for at the worst moment. We wrote the fix for each, and a skill that does it for you.
Works on your laptop. Production disagrees.
A secret ended up in git, a bundle or a public repo.
An open endpoint, a loop or no cap ran up a bill.
The agent said it works. The tests say otherwise.
Each one starts with the emergency you have right now, then makes sure it can't happen again. You can tell what each one does from its name.
About to deploy? It finds what will break in production before your users do: localhost URLs, missing env vars, files the server can't write, imports that only work on your laptop, and 30+ more.
Already broke? Paste the error: it finds the cause, then checks what will break next.
It finds every secret in your code, git history and the JavaScript you ship to browsers, and says how bad each one is.
Then it walks you through revoking and replacing it, with exact steps for 19 providers.
It finds the uncapped calls, loops and open endpoints that caused it, and adds the limits.
Then it gives you a spending-cap checklist for each provider in your code.
It runs your tests and build and shows the real output.
If anything fails or breaks a project rule, the agent has to say so and fix it.
/ship-check runs every check and gives one verdict: safe to ship, or what will break, leak or cost you.
When something is on fire you don't want an essay. Every skill gives you the verdict first, then only what you need to act.
One line: safe to ship, or what will break, leak or cost you. Detail comes after.
Finding, Evidence (file and line), Impact, and Fix. Nothing vague like "consider reviewing your config".
The skill reads your code. It doesn't touch it.
Every fix is shown as a diff first and applied only when you say yes.
Your hosting dashboard, live database or real traffic. It tells you exactly where to look instead of guessing.
Secret OpenAI key behind a browser prefix: NEXT_PUBLIC_OPENAI_KEY
app/chat/page.tsx:12 and .next/static/chunks/…js (sk-p…9f2Q)
Anyone who loads your site can read this key and run requests billed to you.
Rotate the key first, then move the call to a server route: - const key = process.env.NEXT_PUBLIC_OPENAI_KEY + const res = await fetch('/api/chat', opts)
Each skill was scored on blind tests written by someone who never saw it, next to what you'd use without it.
Blind test, JavaScript and TypeScript projects: 33 planted bugs
Blind test: 35 planted leaks
7 blind bill-spike emergencies (repo + usage data)
"Fully solved" means the right cause, the right stop-the-bleeding steps, correct provider cap steps and a working fix. Prove It Works is free, so it has no scoreboard here: it makes your agent verify its work every time, not only when you remember to ask.
Each skill is a folder with a SKILL.md file and small scripts that use only Python's standard library. Nothing else to install.
~/.claude/skills/, or .claude/skills/ inside your projectFull install guide for Claude Code, Claude.ai, Cursor and Codex
Your agent picks the skill up on the next session and loads it when your problem matches.
Paste the error or the alert, or say what you're about to do. You get the verdict first.
Buy the one you need right now, or everything at once.
Fix the one problem you have today. Yours to keep.
/ship-check runs every check and gives one verdict before you ship.
Everything, plus every new skill.
Launch offer: 100% off. Every skill and the bundle are free to download right now, and what you download is yours to keep. Checksums
Each page has the exact error, the cause in plain words, the fix by hand, how to stop it coming back, and links to the official docs. Free to read.
Claude Code, Claude.ai, Cursor, Codex and other agents that read SKILL.md files. Each skill is a folder with a SKILL.md file and small Python scripts. Broke After Deploy is built for JavaScript and TypeScript projects (Next.js, Vite, React, Node, Express).
Yes. The checks use only Python's standard library and make no network calls, so they run without internet. Your agent still needs its own connection to its model. Key Leaked never tests whether a key is live; it assumes an exposed key is live and tells you to rotate it.
During the launch offer everything is free, so there is nothing to refund. Tell us what went wrong and we will fix the skill. Once paid plans start, refunds are no-questions.
Not through TechFix. In Claude Code, Cursor, Codex and other local agents, the scripts run on your machine, make no network calls, send no telemetry and need no account. Key Leaked never prints a full secret, in chat or in any file. Your agent reads the report the same way it reads your code. In Claude.ai, skills run in Claude's own sandbox on the files you give it.
Not without asking. Every skill is read-only by default. Fixes come as diffs and are applied only after you approve them. With your OK, a skill adds one line per run to .techfix/ledger.md so you can see what it caught; delete or gitignore it any time. Key Leaked never rewrites your git history; it shows you the commands.
You can, and for many things it does fine. On the same blind test, plain Claude caught 94% of the planted deploy bugs with 11 false alarms; Broke After Deploy caught all 33 with none. The skills add scripted checks, a list of known causes, and provider steps with sources, so the answer doesn't depend on what the agent remembers that day.
Anything outside your code: your hosting dashboard, live database, real traffic, provider dashboards and invoices you haven't pasted. Each report says which of these matter for your problem and exactly where to check them.
Leave your email and we'll tell you when new skills land.