kosolapov.com
All agent resources
FOR AGENTS / SKILL KIT

CT toolkit: one ZIP for Codex and Claude Code

Give the archive to your agent and run a public CT demo. Includes a coordinating skill, the Slicer skill, the TotalSegmentator skill and CT-CHAT/Merlin templates.

Original resource · Igor KosolapovVersion 1.0.0 · Codex · Claude Code
FIRST RUN

How to install

  1. Download CT-AI-Kit.zip into your Codex or Claude Code project folder.

  2. Send the installation prompt. The agent extracts the archive and reads START-HERE.md.

  3. The agent installs the project skill, checks the environment and downloads dependencies for the selected demo.

  4. After execution, open report.html or overlay.png; the mask and log are saved alongside.

Download the ZIP kit
USE CASES

When to use it

  1. Start from one archive in Claude Code or Codex

  2. Produce a mask and segmentation overlays for a public CT sample

  3. Continue in Slicer or prepare a CT-CHAT/Merlin run

What you’ll get

After a successful run: a liver mask, three slice overlays, an HTML report and a run log. The agent checks the environment and reports a specific blocker if execution is unavailable.

For example

Unpack CT-AI-Kit.zip, read START-HERE.md, install the kit for the current agent and run the public demo. Show the mask overlay and HTML report.
EDITORIAL NOTES

Before you start

The complete toolkit from the article is packaged together. Download the ZIP and give it to your agent; no need to gather the individual cards.

  1. Save the archive in the project folder, or attach it to the task if ZIP attachments are supported.

  2. Send the installation prompt from this page to the agent.

  3. Let it check the environment and install the needed dependencies in an isolated environment.

  4. Open the generated report; results exist only after actual execution.

Walkthrough and step-by-step guide

Included

  • The coordinating ct-ai-kit skill and English startup instructions.
  • Unmodified Slicer skill sources and selected TotalSegmentator skill/MCP files with Apache-2.0 licenses.
  • Original CT-CHAT and Merlin planning templates in Russian and English.
  • Environment check, installer, CLI demo and PNG/HTML report generator.

What you need

  • Codex or Claude Code with local file access and command execution.
  • Compatible Python 3.10+, Git, internet and disk space for dependencies and weights; the agent checks the environment.
  • A compatible PyTorch runtime: GPU/MPS when available, or a slower CPU route.
How it works and what was tested
  • The ZIP contains all toolkit instructions and helper code; Python, 3D Slicer, full libraries and weights are separate per-workflow installations.
  • The default demo uses TotalSegmentator CLI on one public sample. CT-CHAT/Merlin do not automatically run alongside it.
  • The installer does not change global settings or connect MCP; MCP requires a separate setup step.
  • Model output is not a clinical assessment. Runtime and availability depend on hardware and permissions.

All 22 upstream files were hash-verified; 8 packaging, installation and failure-handling tests passed, including both skill layouts. Report rendering was tested on synthetic volumes. Full model execution through this ZIP, Slicer MCP and execution in another Claude Code/Codex account were not tested.

SHA-256: 65cdd2374ef5d5cbb517b25dafcf8cd9c01ead52ff6a3e9efbbe6bf49fb8c8b9