Small, local tools that make Claude catch its own bugs, hold your voice, and stop burning tokens. No accounts, no cloud. You paste a prompt and it installs itself.
🎬 Sound on. And yes: this video was made with one of the tools.
These tools run inside Claude Code and hand you a brief. Here is the actual output, toggle between walking in blind and walking in with the recon.
Prefer to watch? The same brief in 25 seconds.
One page on any fund before you pitch: their real thesis, where you fit, and the objection that kills the deal.
Reads the Reddit threads where your buyers complain and hands you their exact words, top pains, and the gap nobody fills.
Turns a failure into a real root-cause analysis, inside the same session. Stop shipping the same bug twice.
Catches the words and patterns that make writing sound like a robot, before you hit publish. Runs locally, zero tokens.
Teaches Claude your voice and holds it across every draft. New: define any other voice too, a client's, boardroom formal, anyone you like.
Brakes that stop Claude from looping and running up your bill. Catches 11 runaway patterns. Two-minute install.
Turns your rough idea into a spec Claude follows, so it builds what you meant, not what it guessed.
Stop running on willpower. A system that holds the plan, tracks the steps, and hands you the next action.
A backend that remembers your project, decisions, and to-dos across sessions, so you stop re-explaining everything.
Claude negotiates your lease over email. Voss tactics, a calculator that catches the math games, every reply drafted.
Type a sentence, get the launch video. Three production recipes plus the worked templates. The demo video here was made with it.
You can write these hooks yourself. I did, that is where they came from. Here is the honest tradeoff before you decide.
| Write your own | A claudedaddy kit | |
|---|---|---|
| Time to working | A weekend per tool, more to harden it | Two minutes, paste a prompt |
| Tests | You write them, or you skip them | 35 to 210 tests, shipped green |
| Edge cases | You find them in production | Already found and handled |
| Install | You wire settings.json by hand | Done for you, no terminal, no git |
| Updates | You track every Claude Code change | Patched as the platform moves |
| Cost | Your weekend | $29 to $49, one time |
If you like building infrastructure and have the weekend, write your own. The hooks are not magic. The value is the tests and the edge cases I already hit. If you want it working today, the kit is the shortcut you would have built anyway.
.claude folder, wiring settings.json for you.claudedaddy is a store of small, local tools for Claude Code. They make Claude catch its own bugs, hold your writing voice, and stop burning tokens. The gates are real code, not prompts. Each installs by pasting a prompt into Claude Code, with no terminal or git.
No. Each tool installs by pasting a single prompt into Claude Code, which wires it in for you. You need a paid Claude plan. One free tool has an optional terminal step for its Reddit connection.
Some are. The VC Prep Brief and What Your Customers Actually Want are free. The kits are one-time purchases from $29 to $49, plus a memory backend at $39 a month. Two bundles collect the kits for less.
The one-time kits run from $29 to $49. Give Claude a Memory is $39 a month. The Builder Bundle is $99 and the Complete Bundle is $129. The VC Prep Brief and What Your Customers Actually Want are free.
Yes. The tools install into and run inside Claude Code, which requires a paid Claude plan (Pro, Max, Team, or Enterprise). The Code tab does not open on the free Claude plan. No other subscription or account is needed.
No. claudedaddy is an independent set of tools built by Assaf Kipnis. It is not affiliated with, endorsed by, or operated by Anthropic. The tools extend Claude Code; they are not an official Anthropic product.
Open Claude Code in the desktop app, pick a folder, and paste the short install prompt that comes with the download. Claude Code unzips the files and wires the hooks or plugin into your .claude folder. You only approve the steps. No terminal, no git.
Root Cause for Claude is a $29 Claude Code plugin that turns a failure into a real root-cause analysis. It separates the surface cause from the structural cause, forces multi-factor analysis, and demands verification you actually ran. It stops you fixing the same bug twice.
The AI Writing Detector is a $29 local tool that catches the words and patterns that make writing sound like AI before you publish. It checks machine words and cadence, plus a substance check that fires when a draft has no real witness or fact. Zero tokens.
Claude Code Budget Guard is a $39 hook kit that stops Claude Code from looping and running up your token bill. It runs before every tool call, catches 11 runaway patterns like retry loops and agent-spawn storms, and blocks the turn. No model in the loop, so it cannot be argued out of its verdict.
Yes. Every paid tool is sold through Gumroad with a 30-day money-back guarantee. If a tool does not do what the page says, reply to the receipt email and Assaf reads it.
Skills are instructions the model reads and can ignore. The paid kits are hooks: real code that runs outside the model and blocks the turn when Claude drifts, burns tokens, or breaks your voice. Directories list skills; they do not ship enforcement stacks, the one-paste install, or outcomes like a negotiated car lease. Half this line is free on purpose, and the free tools belong in those directories.
Assaf Kipnis, a security practitioner with 12 years in threat intelligence at LinkedIn, Google, Meta, and ElevenLabs. He runs these tools on his own Claude Code and builds the same discipline into team repos as consulting.
Want one wired into your repo, or a heavier spec-and-review pipeline for a crew running agents in production? That part is the consulting. We talk, I scope it, you get it built.
I'm Assaf. I spent 12 years in threat intelligence, at the kind of places where the same failure showed up on four teams and got fixed four times. The learning never stuck.
I build tools that make it stick. These are the ones I run on my own Claude Code, every day. The gates are real code, not prompts, so the discipline holds even when the model would rather hand you something tidy and wrong.
If you want that discipline wired into your team's repo, that's the consulting. Everything here is the free and cheap version of the work I do paid.