Draft a scientific workflow diagram

Open your own project in Codex or Claude Code, with permission to read and write files and run commands. An ordinary web chat may not be able to install a Skill.

• Codex or Claude Code with permission to read and write project files and run local commands. A chat page without file tools cannot install this skill.
• Python 3.10+, requests, network access, and an OpenRouter API key with available credit. Generation and AI review are separate paid requests.
• An agreed cost estimate, call plan, and maximum budget before any paid request. The helper has no built-in monetary spending cap.
• Approved content for sending to the external service. Configure the named API key through the local environment, without putting it in the diagram prompt.

These requests run in your assistant's host environment. This page provides guidance; it does not run AI tasks or collect your project files.

Copy this request into your project conversation. Ask the assistant to review the original Skill and explain any required setup before using it.

I am using Codex or Claude Code with filesystem and shell tools. Install Scientific Schematics from https://github.com/K-Dense-AI/scientific-agent-skills at commit 593377bb048752abb0152deab9fbb1083a6087df, folder skills/scientific-schematics. Read SKILL.md and retrieve the complete skill with its references, scripts, and assets when present, not just the instruction file. Preserve the original repository LICENSE.md and K-Dense Inc. attribution. Use the current project's skill directory supported by this host, following existing project configuration. Do not overwrite an existing skill; explain conflicts first. Check dependencies and explain proposed environment changes before installing them. Tell me the installation path and how to invoke it, then wait for my files and task. Do not make any paid API calls during installation or setup. Before any future paid generation, show the proposed models, call count, current price estimate, and maximum budget, and wait for my explicit confirmation.

Once installation is confirmed, replace the example paths and details with your own files, then send this request in the same project.

Use the installed scientific-schematics skill to plan a simple four-box workflow: Sample preparation → Measurement → Data analysis → Report. Preserve my original notes and any existing figures. First show the proposed labels and connections, the generation and review models, the planned number of calls, an estimate using current provider prices, and a maximum budget for my confirmation. Stop and wait for my explicit budget confirmation before making any paid API call. Only after confirmation, run at most one generation-and-review iteration for the approved content; do not retry or start another paid call beyond that plan without renewed confirmation. Save the PNG and review log in a new output folder, then inspect every label and arrow with me. Deliver a draft for manual checking, not an editable vector file or scientific certification.

• Raster PNG drafts, with numbered versions when more than one generation is explicitly approved.
• A JSON review log recording available critiques, scores, errors, and the final selected image; all results still require manual verification.

Ask your assistant where each output was saved and how it was checked. Compare the result with your original inputs before you use or share it.

• Source and documentation checked on 2026-10-20; complete runtime workflows have not been tested here.
• This is a paid online workflow: prompts and generated images are sent to OpenRouter and the underlying providers. Their charges and data policies apply.
• The output is raster PNG only, with no editable vector path or DPI control. Converting it to PDF or enlarging pixels does not create vector detail.
• AI review scores do not establish scientific validity, accessibility, or publication acceptance. Manually verify labels, units, counts, topology, and arrow directions.
• The script can make a second generation by default and has no monetary budget cap. Confirm the budget first, explicitly limit the approved run to one iteration, and do not replay paid calls automatically.

A third-party open-source selection by AimatraLab. The Skill belongs to its original author; source links go directly to their repository.

Original author

K-Dense Inc.

Repository

https://github.com/K-Dense-AI/scientific-agent-skills

Selected version

1.10

Fixed commit

593377bb048752abb0152deab9fbb1083a6087df

License

MIT

Documentation checked

2026-10-20