UiPath Agent Skills

[!NOTE] Work in Progress — This repository is under active development. Skills are being added and refined. Contributions, feedback, and ideas are welcome! See Contributing below.

UiPath Agent Skills give AI coding agents the domain knowledge to build, run, test, and deploy UiPath automations and agents — directly from your development environment. Each skill is a self-contained package of instructions and resources that teaches your coding agent how to perform a specific UiPath task.

Quick Start

Prerequisite: Node.js (LTS) is required — it includes npm.

npm -g install @uipath/cli
uip skills install

uip skills install finds the AI coding agents installed on your machine and installs the skills for all of them, into each agent's directory, ready to use. If it can't find any agent, it asks which one to target. To install for just one agent, pass --agent <name> (e.g. --agent claude).

brew install node

winget install OpenJS.NodeJS.LTS

curl -fsSL https://fnm.vercel.app/install | bash
fnm install --lts

Try a skill without installing anything

To evaluate a single skill before adopting the whole set, use the skills CLI — no UiPath CLI and no project changes required:

npx skills add UiPath/skills --list              # browse the catalog
npx skills use UiPath/skills@uipath-rpa | claude # run one skill; nothing is written to your project
npx skills add UiPath/skills --skill uipath-rpa  # install just that one skill

For ongoing use, prefer uip skills install above. It detects every coding agent on your machine and keeps skills updated as they change; npx skills add installs a point-in-time copy that you must update yourself with npx skills update.

Skill Catalog

The repository ships skills covering authoring, platform operations, and diagnostics for UiPath automations.

Authoring

Solution & Planning

Platform & Operations

Diagnostics & Feedback

Lifecycle Status

Every skill's maturity is tracked in assets/skill-status.json — the source of truth. The table below is generated; refresh it with python3 scripts/check-skill-status.py --write-readme.

Status legend: - Stable — Stable, production-ready surface; safe for production. - Preview — Not yet stable; may be broadly available or gated/allowlisted, and surface and behavior may change. - In-development — Skill itself is incomplete or unstable; coverage is partial.

Agents

Multi-Tool Support

This repository works with Claude Code, Google Gemini CLI, OpenAI Codex CLI, and Cursor IDE.

Claude Code

This repository works as a Claude Code plugin. Install skills via uip skills install (the recommended method above) — this keeps skills updated automatically.

Avoid manual plugin installation (e.g., claude plugin marketplace add / claude plugin install). Manually installed plugins must be updated manually, and you may miss skill updates and fixes.

If you've already installed manually, uninstall and re-install via uip skills install to switch to automatic updates.

Reduce permission prompts

By default, Claude Code prompts for approval on every uip command, and a realistic Flow or RPA build runs 25+ distinct subcommands. Claude Code plugins cannot ship permission allowlists declaratively, so run this once to install a curated allowlist of safe read-only commands (registry lookups, validation, local scaffolding) while keeping prompts for side-effectful operations (login, debug, publish):

/uipath:install-permissions

The command prints the recommended JSON and offers to merge it into your settings (~/.claude/settings.json for global, or project-local .claude/settings.local.json). You will also see a one-line nudge at session start until an allowlist is installed.

Google Gemini CLI

Gemini CLI is supported by uip skills install. If the Gemini CLI is on your PATH, it's detected automatically and skills are wired up. If no agent is detected, pick Gemini CLI when prompted.

OpenAI Codex CLI

This repository is configured as a Codex CLI skill provider. The AGENTS.md file (symlinked to CLAUDE.md) provides project instructions, and skills are discovered via .agents/skills/ (symlinked to skills/).

This repository also includes Codex plugin metadata under .codex-plugin/. The Codex plugin manifest exposes the same skills/ directory and shared session hooks.

Install with Codex CLI:

codex plugin marketplace add UiPath/skills --ref main
codex plugin add uipath@uipath-marketplace

The marketplace entry currently uses a plugins/uipath symlink so Codex can load the repository root as the plugin root; remove it once openai/codex#17066 is resolved.

Windows users: This repo uses git symlinks. Clone with symlinks enabled: bash git clone -c core.symlinks=true https://github.com/UiPath/skills If you've already cloned without symlink support, re-enable and re-checkout: bash git config core.symlinks true git checkout -- .

Cursor IDE

Project rules are provided in .cursor/rules/ and are automatically loaded by Cursor.

Contributing

Contributions are welcome! Whether it's a new skill, a bug fix, or a documentation improvement — we'd love your help.

See CONTRIBUTING.md for the full guide, including: - Repository structure and architecture - How to add a new skill (folder layout, SKILL.md format, frontmatter) - Naming conventions and quality checklist - Pull request process and branch naming

Quick version:

1. Fork this repository
2. Create a feature branch (feat/add-<skill-name>)
3. Add your skill under skills/uipath-<name>/ with a SKILL.md
4. Submit a pull request

For questions, ideas, or feedback, please open an issue.

Running coder-eval on demand

Run Coder Eval — GH-hosted workflow that runs coder-eval against the skills task tree (tests/tasks/...). Use for ad-hoc single-task verification or a folder-scoped sweep on the same infra as the nightly cron. Trigger from the Actions tab → "Run workflow" → fill in task_globs. Definition: .github/workflows/run-coder-eval.yml.

Resources

• UiPath Documentation
• UiPath Community

License

MIT