Independent directorynot affiliated with xAI
Catalog refreshed Nobody pays to be listed or ranked
9,075 templates from 4,179 creators, each showing what it connects to, whether it asks first and how often it runs. “Checked” means the link was reachable — not a safety rating.
The job here is designing and assembling Grok Bots, with a quality bar aimed at how the Japanese reads. Anything a bot will show to a human is checked for natural phrasing rather tha…
Hand it a strategy or a plan and it argues with the framing instead of nodding along — naming the groupthink, challenging the premise, and laying out what each option really costs. I…
Turns accepting a patch into a contest. You state the property a fix must uphold, one agent writes the change, and separate agents try hard to defeat it. The final gate reruns your r…
Running a crowd of assistants means paying for all of them at once, and the invoice rarely says which one was careless. This gives each persona its own allowance, keeps the skills it…
Codes against a fixed list of acceptance criteria, and for each one it supplies a test that demonstrates the behaviour instead of an assertion that claims it. Before anything is call…
Every pull request gets two readings: one that assumes good intent and looks for design trouble, and one that hunts for what an attacker or race condition could break, prototype poll…
Meant for accounts where several bots share one allowance. It watches how much capacity remains, and passes that figure to the other agents so they can slow down before the account r…
A steganography bench with two directions. It can tuck a private message inside a picture so that only someone holding the passphrase can read it back, and it can inspect pictures, s…
Talks through self-hosted proxy server work in Simplified Chinese, from picking a node to per-protocol deployment and the client-side configuration that connects to it. Covers kernel…
Works as the hiring desk for a fleet of assistants: you pick the role you need from a single list, and it reads the specification over for problems before anything is created. It the…
Name Vandenberg, the Cape, Starbase or any other pad, and it reads public forecasts and model output to grade the coming 72 hours for a launch commit. Each day lands on a three-step…
Reviews websites and repositories you control and reports findings ranked by severity, each one backed by the evidence that led to it. It stays on the defensive side: no proof-of-con…
Point it at a URL and it scores the page's WebMCP readiness, then explains in plain language what is missing and why it matters. What comes back is a short written brief a site owner…
Scoped strictly to applications you own or are authorized to test. It works a structured checklist of the usual break-in routes — authentication weaknesses, access-control gaps, inje…
A design partner for the phase before programming starts. It works a concept into a core loop and the hook that brings players back, then hands over a brief a developer or a coding a…
Treats a technical sketch as a piece bound for readers rather than one ticket among many, taking a rough engineering drawing and refining it until it holds up on a printed page. Each…
Produces a complete track from a style note and a set of words, running the music model on your own Apple-silicon hardware instead of in somebody else's cloud, and returning a high-s…
Describe the job and it works the cut like a shop assistant: which tool, what surface speed, how deep, how the stock should be held, and where the fixturing will fight you. It finish…
Paste in another bot's system prompt and it reads it like a strict editor: what is the one job this prompt exists to do, and what is padding that job? It flags duplication, vague wor…
Send it a link to a WeChat official-account article and it returns the full piece as tidy Markdown and JSON: headline, outlet, publish time, body text, and images all carried through…
A bot's configuration and everything it has learned sit inside the account, with no obvious way to get them back once something breaks. This one copies both out to a private reposito…
Tell your computer what to do from another room and it listens. Routine commands like muting run as a local script so almost nothing reaches the AI, while bigger asks go to the bot a…
A head bot that runs your team of specialist bots: it keeps your priorities, routes work to the right bot, holds the standards, and only escalates the decisions that need you. Built…
Keeps your Grok Bot plugins and connectors lean and working. Audits what's installed, catches broken or signed-out connectors, flags duplicates and risky tools, and sends a short wee…
Wraps a command-line build harness in a bot, so app work, coding and deep-research jobs can run hard on a hosted agent machine instead of occupying your own terminal. Useful when a t…
A designer-and-builder for simple phone web apps, aimed at tools that should work without a manual. Screens carry obvious labels and touch targets sized generously, the way shop-floo…
Sends you one short note whenever a T3 Code, Claude Code, or Codex thread finishes, stalls, or errors, so you know what happened and what needs you. Works with the open-source t3-don…
A read-and-report security teammate for individuals and small teams. It triages GitHub security alerts, scans for leaked secrets, checks your sites'…
Turns your GitHub month or year into one Spotify-Wrapped-style card plus 3 tweet drafts you can post yourself. Private work only shows as totals, and repo names never appear.
Turns all your Grok Bots into a tap-through office building: floors by job, each bot's status, usage and what's waiting on you. Read-only, and it refreshes only when you ask.
Hand it the link or the raw setup of a bot you are thinking of adding, and it works through what that bot is actually wired to do. It pulls out the parts worth knowing about and clos…
For indie developers, review day is where releases quietly die. This one reads your app or extension against the review rules of the Apple Mac store, Setapp and the Chrome Web Store,…
Breaks a coding job into bounded tasks and hands each one to a sandboxed coding agent running on its own machine, with Kilo and OpenCode as the supported workers. It then reads the d…
Modelled on the hard-edged product discipline long associated with Apple. It presses you to state what a user actually hires your product to do, where the narrow entry point sits, an…
A local dispatcher that farms your bot's programming work out to the command-line coding agents you already pay for, naming Claude Code, Codex and xAI's Grok Build among them. Every…
Runs an app or game through two passes and reports what each one found. The first pass plays it like a young child seeing it for the first time, catching whatever confuses a newcomer…