AI coding, games & tech tools ✦ We build and ship small, useful products: prompt packs, templates & dev tools. Start with our Vibe Coding Starter Pack 👇
Joined October 2026
- Tweets583
- Following20
- Followers1
- Likes8
Pinned Tweet
Vibe Forge Labs is live ⚡
We build and ship small, useful products for AI coding, games & tech tools.
First drop: the Vibe Coding Starter Pack. 50 prompts, a cheat sheet & 3 spec templates for Cursor, Claude Code, Replit, Bolt & v0. $9.
vibeforgelabs.gumroad.com/l/…
TikTok LIVE game tip: fire every gift trigger once offline before you go live. Rose, Finger Heart, Galaxy. Watch each command actually run in-game. A trigger that fails silently on stream kills the moment.
78 tested Minecraft gift commands: vibeforgelabs.gumroad.com/l/… #TikTokLive
Next Fest ends in ~23h. Before your demo leaves the front page:
1. Pin a "demo stays live" post
2. Add your launch month to the short description
3. Note which post caused each wishlist spike
Last day for 40% off the Steam Launch Kit: vibeforgelabs.gumroad.com/l/… #SteamNextFest
Not reading every diff is fine. Not knowing which diffs are risky isn't.
Read every line when it touches auth, payments, migrations, deletes, secrets or CI. Skim the rest and let tests gate it.
A hook can flag risky paths before you approve: vibeforgelabs.gumroad.com/l/… #ClaudeCode
5h Claude limit vanishing early? Measure before blaming your prompts.
Put context % and plan usage in a status line, run the same task in Claude Code and in your SDK app, then compare the meters.
The Mods Kit's context-meter mod does it: vibeforgelabs.gumroad.com/l/… #ClaudeCode
Next Fest ends Oct 26. Before players move on: put your launch date and wishlist ask on the demo's last screen, pin your best clip, and reply to every review.
Last hours of 40% off our Steam Launch Kit with NEXTFEST: vibeforgelabs.gumroad.com/l/… #gamedev
Still vibe coding if you read every diff? Yes, but you don't need to read every line. Add a gate: the agent explains each changed file, names the test that proves it, and can't push to main.
5 ready-to-run Claude Code mods for that: vibeforgelabs.gumroad.com/l/… #ClaudeCode
Tests your agent wrote aren't proof. They check what it built, not what you meant.
Before it codes, write 3 acceptance cases in plain English: the input and the result you expect. Have the agent turn them into tests first, then code to pass. #vibecoding
vibeforgelabs.gumroad.com/l/…
#ClaudeCode tip for vibe debugging: don't say "fix the bug." Have the agent write a failing test that reproduces it, show it failing, then fix until it passes.
You review one test, not a whole diff, and the bug can't sneak back.
Hooks + guard rails: vibeforgelabs.gumroad.com/l/…
#ClaudeCode tip: never let an agent merge its own PR.
PreToolUse hook on Bash: if the command matches gh pr merge or git push to main, exit 2 with "needs a human". The agent opens the PR; you click merge.
Guard hook + 4 more mods: vibeforgelabs.gumroad.com/l/…
Claude Code container restart ate your work? Add a Stop hook that commits and pushes to a wip/ branch after every turn. Squash before you merge. Costs nothing, saves hours. More guard-rail hooks like this in the Mods Kit: vibeforgelabs.gumroad.com/l/… #ClaudeCode
Why do vibe-coded sites all look the same? The AI gets zero taste input.
Fix: before the first prompt, put 5 design tokens (2 colors, 1 font, radius, spacing) and 2 reference screenshots in CLAUDE.md.
3 kits for shipping with AI, $24: vibeforgelabs.gumroad.com/l/… #vibecoding
Your Claude Code setup shouldn't live on one laptop.
Commit it with the repo:
- CLAUDE.md (rules + context)
- .claude/settings.json deny rules
- PreToolUse guard hooks
New teammate clones, gets the same guard-rails.
Hooks + templates: vibeforgelabs.gumroad.com/l/… #ClaudeCode
Next Fest ends Monday. Before your demo slides off the front page:
1. Post a thank-you with your wishlist count
2. Turn the top 3 feedback points into a dated devlog
3. Keep the demo up with a wishlist end screen
Templates, $9 till Mon: vibeforgelabs.gumroad.com/l/… #SteamNextFest
If your coding agent can edit .claude/settings.json, it can loosen its own rules. Add a deny rule like Edit(.claude/settings.json), plus a PreToolUse hook that blocks writes to it. Prompts ask nicely. Hooks enforce. #ClaudeCode
The hook setup I use, with copy-paste configs: vibeforgelabs.gumroad.com/l/…
People are letting Claude Code "fix" their TVs. One quietly disabled 7 apps nobody asked about.
Before it touches anything real:
1. PreToolUse hook blocking destructive commands
2. Allowlist exact targets
3. Log every tool call
Guard hooks: vibeforgelabs.gumroad.com/l/… #ClaudeCode
Steam Next Fest ends Monday. Before the traffic fades:
1. Put your wishlist ask in the first two lines of the demo page
2. Post a devlog on what players told you
3. Announce your post-fest update date
Steam Launch Kit, NEXTFEST code to Oct 26: vibeforgelabs.gumroad.com/l/… #indiedev
A Supabase app just lost 35k profiles through a view that skipped RLS.
Before you ship:
1. Create views with security_invoker = true
2. Revoke anon grants you don't use
3. Query as anon: expect nothing
Full checklist, 10% off: vibeforgelabs.gumroad.com/l/… #vibecoding
Claude Code just got sandbox + MCP permission fixes. Re-check your guard hooks:
1. Does PreToolUse still block rm -rf and .env reads?
2. Do MCP tool calls hit the hook too?
3. Run one fake bad command and confirm it's denied.
Tested hooks: vibeforgelabs.gumroad.com/l/… #ClaudeCode
Weekend build plan for Claude Code or Cursor:
Sat: 5-line spec, build the core loop
Sun: security pass (keys out of git, RLS on, agent blocked from .env)
Then a guard hook so the agent can't rm -rf
Prompts, security + mods kits, $24: vibeforgelabs.gumroad.com/l/… #vibecoding