Skip to main content
The Ionworks Agentic Toolkit gives coding agents the context they need to drive Ionworks Studio end-to-end: ingest cycling data, organize cell characterization, run simulations, and structure research projects — without hand-holding on endpoints, schemas, or conventions. Each skill is a single SKILL.md file with YAML frontmatter. Any agent that understands the skill convention — Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Anthropic SDK agents, and others — can load them.

What you get

install

Installing the SDK, configuring IONWORKS_API_KEY, verifying the install, upgrading, troubleshooting auth and proxy errors.

discover-api

Always call client.capabilities() and client.schema(name) first so later steps don’t guess endpoint shapes.

process-data

Convert raw cycler data (MATLAB, CSV, JSON, vendor exports) into the platform’s parquet + JSON hierarchy or the portable BDF format.

validate-data

Validate processed data against the platform’s authoritative schemas and strict measurement checks before upload — schema, parquet, and domain validators in one gate.

upload-data

Upload time-series, file, and properties measurements. When to use each measurement type.

manage-cells

Create and query cell specifications, instances, and measurements. Parent/child relationships across the cells domain.

run-simulations

Run electrochemical simulations with UCP protocols, design-of-experiments sweeps, and retrieve results.

sim-results

Fetch and analyse simulation output — time-series and step DataFrames, SOC, DCIR, step identification, plotting, and the per-step capacity-reset caveat.

run-simple-pipelines

Fire-and-forget single datafit or validation runs via SimplePipeline — submit a config, poll for completion, retrieve the result.

manage-projects

Organize work across projects, studies, models, parameterized models, and optimizations.

build-data-report

Build a comprehensive multi-section data analysis report (rate capability, DCIR, OCV, GITT, entropic, aging, gap analysis) for cells whose measurements live on the platform — markdown + PDF.

Quick install — paste this into your coding agent

Drop the block below into Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, or any agent with shell access. It auto-detects which agent it’s running in, picks the right install path, verifies the SDK is set up, and tells you what’s missing.
The agent fetches the canonical README on every run, so install instructions stay in sync as new agents are supported.

Manual install paths

If you’d rather drive it yourself, pick the path that matches your agent.

Any agent (auto-detect)

The fastest path uses the skills CLI from vercel-labs/skills. It auto-detects your agent and installs to the right directory:
Add --all to install every skill non-interactively, or --skill <name> to install just one. npx skills update -g keeps the install in sync with new releases. This path natively covers Cursor (skills land in ~/.cursor/skills/) and GitHub Copilot for VS Code and the CLI (Copilot auto-discovers SKILL.md from .agents/skills/ and .github/skills/ — no extra config).

Claude Code

Native plugin install — also brings in the slash commands and the SessionStart skill-index hook:
Verify with claude plugin list — the entry should show Status: ✔ enabled.

Codex

Codex needs a marketplace directory that references one or more plugins. This repository ships .codex-plugin/plugin.json but not a marketplace manifest, so install via a small local scaffold:
1

Clone the public skills repo

2

Create a marketplace directory that points at it

3

Register the marketplace and install the plugin

Verify with codex plugin list — the entry should show (installed, enabled).

Gemini CLI

Gemini CLI picks up extensions that ship a gemini-extension.json at the repo root. Install via the Gemini CLI extension command:
After install, every skill is loaded into every Gemini session via GEMINI.md as passive context. Ask Gemini to use an Ionworks skill by name — e.g. “Use the discover-api skill” — and the loaded context will guide it.

Claude Agent SDK / API

Point your agent at any SKILL.md file directly:

Manual

Clone the repo and copy any skills/<skill-name>/SKILL.md into your agent’s skills or context directory. The frontmatter description acts as the trigger; the body is general-purpose guidance the agent reads when activated.

Quick start

1

Install the skills

Use the paste-in prompt above, or npx skills add ionworks/ionworks-skills -g.
2

Install the Python SDK

In the environment your agent will execute code in:
See the Python API client page for details.
3

Export your API key

Get one from your account settings, then:
4

Verify with the install skill

Ask your agent something like “Use the ionworks install skill to verify my setup” — it will invoke install, then discover-api, and confirm the connection.

Requirements

  • Python 3.12+ in the environment the agent will execute code
  • ionworks-api — the Python SDK
  • Optional: ionworksdata for process-data workflows
  • Optional: ionworkspipeline for pipeline-based parameterization (licensed — contact Ionworks)

Model capability

The skills assume the agent can read Markdown, follow multi-step procedures, call a Python SDK, and handle JSON/parquet data. They’ve been authored and tested against Claude Sonnet / Opus and should work with any comparable frontier model. Smaller local models may struggle with the multi-step data-processing workflows in process-data and run-simulations.
Source code, full skill text, and releases live at github.com/ionworks/ionworks-skills.