KryllOS Skill comes built into KryllOS. Plug it into the AI you already use, describe a strategy in your own words, and your AI places the blocks in the editor, backtests them and fixes what breaks. You get ordinary blocks back: you read them, change them, and decide when they go live.

Works with

  • Claude CodeAnthropic
  • OpenAI CodexOpenAI
  • Gemini CLIGoogle
  • CursorAnysphere
  • GitHub CopilotGitHub

…and any agent that reads the open Agent Skills format.

Claude Codeyour AI

Buy 25% when the 4h RSI drops under 35, but only while the 12h trend is up. Sell at −2%, at +4%, or after 48 hours, whichever comes first.

KryllOS SkillSKILL.md
  • Every block
  • How the engine thinks
  • The .kryll file
  • The validator
  • The backtest loop
  1. 01Reads the skill
  2. 02Places the blocks
  3. 03Checks the wiring
  4. 04Runs a backtest
  5. 05Hands it back

What changes when your AI has the skill

A general-purpose AI knows what an RSI is. It does not know how KryllOS runs one. The skill is the manual it was missing, written from the KryllOS source code: every rule points to the file that implements it, so your AI can check it instead of taking it on trust.

  • The blocks

    Your AI alone — Guesses block names and settings, and sometimes invents them.

    With the skill — Knows every block in the palette, its settings, its defaults and its traps, like the moving-average block that ships with a 5% gap nobody asked for.

  • The waits

    Your AI alone — Writes a condition that may never come true. The strategy freezes, silently.

    With the skill — Gives every wait a way out, a timeout or another signal, and never builds a loop that spins without pausing.

  • Stop and target

    Your AI alone — Computes them from the price of the signal.

    With the skill — Computes them from the price you were actually filled at, the way a trader would.

  • The testing

    Your AI alone — Hands you a draft it never ran.

    With the skill — Checks the graph, backtests it on your own KryllOS, reads the trades and the logs, and makes sure every exit (stop, target, timeout) actually fires.

  • The result

    Your AI alone — Produces something only it understands.

    With the skill — Writes a file that opens in your editor with no red block, every block labelled, every link one you could redraw by hand.

  • Your version

    Your AI alone — Works from whatever it remembers.

    With the skill — Checks which KryllOS you run first. When the code and the skill disagree, it follows the code.

Works with the AI you already use

KryllOS Skill is written in Agent Skills, the open format that the major coding agents read natively. One skill, the same rules, whichever agent you prefer.

  • Claude CodeAnthropicAnthropic’s agent, in your terminal, your IDE or the Claude desktop app.
  • OpenAI CodexOpenAIOpenAI’s coding agent, in the CLI, your IDE or the ChatGPT app.
  • Gemini CLIGoogleGoogle’s open-source agent, right in your terminal.
  • CursorAnysphereThe AI code editor and its built-in agent.
  • GitHub CopilotGitHubCopilot’s agent mode, in VS Code and on GitHub.
  • And the othersOpenCode, Goose, Amp, Kiro, Junie, Mistral Vibe: any agent that reads the Agent Skills format can use it.

It does not guess. It tests.

An AI that writes a strategy and stops there hands you a draft. With the skill, it goes around the loop a careful trader would: write, check, replay, read the results, fix. You get the strategy after it has survived its first backtest, not before.

  1. Writeblocks in your editor
  2. Checkvalidator · 0 errors
  3. Replaybacktest on real history
  4. Readtrades · logs · every branch
  5. Fixthen around again

What stays in your hands

  • Your exchange keys

    Your AI works on strategies and backtests. Your API keys stay in KryllOS, on your machine or your server, and the skill never asks for them.

  • The go-live button

    The AI builds and tests. Sending a strategy live is something you do yourself, from KryllOS, once you have read what it built.

  • The road your data takes

    Your AI talks to your own KryllOS, on your computer or your server, through a private SSH connection. There is no Kryll server in the middle: your strategy goes to the AI you chose, and nowhere else.

  • The decision

    The skill teaches how to build a rule, not which rule to build. Nothing your AI says with it is investment advice, and a good backtest predicts nothing.

KryllOS Skill, in questions

The short answers. The long ones are in the skill itself.

What is KryllOS Skill?

A skill is a folder of instructions an AI agent loads when a task calls for it. KryllOS Skill is the one that teaches an agent to build, check and backtest trading strategies in KryllOS. It comes built into KryllOS.

Which AI agents does it work with?

Claude Code, OpenAI Codex, Gemini CLI, Cursor and GitHub Copilot, plus any agent that reads the open Agent Skills format, such as OpenCode, Goose, Amp, Kiro, Junie or Mistral Vibe.

Can I use it with ChatGPT?

Yes, through Codex, the OpenAI agent that comes with ChatGPT plans. A plain chat window is not enough: the AI has to reach your KryllOS to write the strategy and run the backtest, and that is what an agent like Codex or Claude Code can do.

Do I need to know how to code?

No. You describe the strategy in plain words. The AI writes the blocks, and you read them in the visual editor, where a strategy reads like a sentence.

Can the AI trade with my money?

No. The skill only builds strategies and runs backtests in your KryllOS. It never asks for your exchange keys, and sending a strategy live stays a button you press yourself.

Can I change what the AI built?

Yes. It is a normal strategy in your editor: move a block, change a setting, delete a branch, or ask your AI for the next change and it edits the same file.

How much does it cost?

The skill comes free with KryllOS. You only pay for your AI, on the plan you already have with its provider.

Get KryllOS. Bring your AI.

The skill is already inside. Download KryllOS, point Claude Code, Codex or your favourite agent at it, and describe your first strategy.

Ready for

  • Claude Code
  • OpenAI Codex
  • Gemini CLI
  • Cursor
  • GitHub Copilot