How uipath/skills works
This repository is a documentation-and-tooling hub for UiPath “skills,” not an application runtime. The core product surface is a catalog of agent instructions that teach coding agents how to work with UiPath products such as RPA, Flow, Maestro, Platform/Orchestrator, Solutions, Agents, and Coded Apps.
The main architectural boundary is between skill content and supporting automation. Skill content lives under skills/ as self-contained packages, each centered on a SKILL.md and usually accompanied by references/, assets/, and sometimes fixtures/. Supporting automation lives in tests/, .github/, scripts/, and hooks/, which validate that the content stays correct and installable.
The repository is designed to be consumed by multiple agent runtimes. The README shows support for Claude Code, Gemini CLI, OpenAI Codex CLI, and Cursor IDE. That integration is mostly file-based: AGENTS.md/CLAUDE.md, .claude/, .cursor/rules/, .gemini/, and .agents/skills/ act as the configuration and discovery layer for those tools.
At the distribution layer, uip skills install is the primary entry point. The repository includes a Claude plugin definition (.claude-plugin/) and plugin metadata, but the recommended path is installation through the UiPath CLI so skills can be discovered and updated automatically. The repo also includes a project-level hook setup script and permission guidance for uip commands, indicating that local agent workflows are expected to invoke the UiPath CLI frequently.
The skill catalog is organized by product capability rather than by implementation code. The major skill domains are:
• uipath-rpa for XAML and coded workflows
• uipath-maestro-flow for .flow JSON authoring and debugging
• uipath-maestro-bpmn and uipath-maestro-case for BPMN/case authoring
• uipath-agents for coded/low-code agent lifecycle
• uipath-coded-apps for web/action app lifecycle
• uipath-platform, uipath-solution, uipath-review, uipath-troubleshoot, and others for platform operations and governance
Each skill is structured as a documentation bundle with a stable internal layout. SKILL.md is the entrypoint; references/ contains the authoritative how-to and domain rules; assets/ provides templates and starter artifacts; fixtures/ holds sample projects or validation inputs; and in some skills .maintenance/ contains link/check scripts that keep the documentation tree coherent.
The repository has a strong agent subsystem as well. agents/uipath-project-discovery-agent.md is used to infer project conventions and generate context files for downstream tools. This is a meta-layer that helps other repos/projects become “skill-aware” by auto-generating instructions such as .claude/rules/project-context.md and AGENTS.md.
The test architecture is substantial and is the main quality gate. tests/tasks/ contains scenario-driven fixtures grouped by skill domain, with YAML task definitions plus Python checkers. These tests cover smoke, integration, E2E, and negative cases. They exercise the actual skill instructions and validate outputs against expected repository behavior, often using mock responses and generated project folders.
A notable testing pattern is domain-specific fixtures paired with executable assertions. For example, the RPA and Troubleshoot task suites include XAML projects, mocked Orchestrator responses, and Python helpers that inspect the generated artifacts. The Maestro flow and BPMN suites similarly use template projects, JSON/BPMN resources, and check scripts to validate structure, wiring, and lifecycle semantics.
The CI/CD layer is implemented with GitHub Actions. Workflows such as validate-skills, smoke-skills, smoke-rpa-skills, lint-tasks, test-helpers, and activation-gate run repository-level checks. There are also automated maintenance workflows for version bumps and PR review, showing that skill content is treated like a release artifact with quality gates, not just static docs.
Repository automation is reinforced by local scripts and hooks. scripts/setup-hooks.sh installs Git hooks, hooks/ contains shell scripts that enforce guardrails such as ensure-uip.sh and skill description validation, and .githooks/pre-commit supports fast local enforcement. These pieces keep authoring conventions aligned before CI runs.
The main data flow in the repo is: author/edit skill content in skills/ → validate structure and wording via hooks/tests/CI → package/distribute via the UiPath CLI and plugin metadata → consume from agent runtimes using the repo’s config files and symlinked entrypoints. External dependencies are mostly UiPath products and services such as Orchestrator, Integration Service, Studio Web, and UiPath CLI, which are referenced heavily in the skill docs and test scenarios.
For an architecture diagram, the most important boxes are: the skill catalog, the multi-agent/tooling integration layer, the test/validation harness, and the GitHub automation boundary. The key arrows are from skill docs to agent runtimes, from tests/CI back to skill docs, and from skills into UiPath platform services during real user workflows.
Main entry points: skills Skill catalog · tests Validation hub · uipath-rpa RPA skills · uipath-maestro-flow Flow skills
Components 22
Skill catalog
Self-contained UiPath skill packages and their documentation bundles.
- Skill catalog · Collection of product-focused skill bundles centered on SKILL.md files.
skills - RPA skills · Authoring guidance for XAML and coded workflows plus activity documentation.
skills/uipath-rpa - Flow skills · Maestro flow authoring, editing, evaluation, and runtime conventions.
skills/uipath-maestro-flow - BPMN skills · Maestro BPMN authoring, validation fixtures, and shared semantics.
skills/uipath-maestro-bpmn - Case skills · Maestro case modeling, stages, tasks, triggers, and variable rules.
skills/uipath-maestro-case - Agents skills · Coded and low-code agent lifecycle, integrations, and fixtures.
skills/uipath-agents - Coded apps · Action and web app lifecycle, SDK patterns, and deployment.
skills/uipath-coded-apps - Platform ops · Orchestrator, licensing, traces, and platform-level procedures.
skills/uipath-platform - Solutions · Solution design, product selection, packaging, and deployment.
skills/uipath-solution - Troubleshoot · Incident analysis, playbooks, schemas, and agent-driven triage.
skills/uipath-troubleshoot - Review · Architecture and implementation review guidance across UiPath products.
skills/uipath-review
Agent integration
File-based discovery and runtime-specific entrypoints for different coding agents.
- Agent layer · Cross-runtime discovery and instruction files for multiple agent tools.
- Claude/Cursor/Gemini · Runtime-specific commands, rules, and settings used by external coding agents.
- AGENTS/CLAUDE · Top-level agent guidance files used as bootstrap instructions.
AGENTS.md - Rules packs · Per-runtime rules and commands that shape task authoring and review behavior.
- Plugin layer · Claude plugin and metadata for packaging/discovery of the skills repository.
.claude-plugin
Validation and delivery
Local hooks, scripts, tests, and CI workflows that keep skills consistent and installable.
- Validation hub · Task fixtures, checkers, tests, hooks, and CI workflows that validate skills.
tests - Task suites · Domain-specific smoke, integration, e2e, and negative task definitions with checkers.
tests/tasks - Hooks/scripts · Git hooks and helper scripts that enforce repository conventions before CI.
hooks - GitHub Actions · Repository workflows for validation, smoke tests, linting, reviews, and maintenance.
.github/workflows
UiPath platform
External UiPath products and services referenced by skills and validation scenarios.
- UiPath services · Orchestrator, Integration Service, Studio Web, Maestro, Agents, and related services referenced by skills.
Other components
- skills repo · Repository for UiPath skill instructions, runtime integration, and validation automation.
Connections 33
- skills repo → Skill catalog: contains · Repository centers on the skill library.
- skills repo → Agent layer: supports · Repository is consumable by multiple agent runtimes.
- skills repo → Validation hub: validated by · Content and conventions are checked by tests and automation.
- skills repo → Plugin layer: packages · Repo includes plugin metadata for distribution and discovery.
- Skill catalog → RPA skills: includes · Major RPA skill package.
- Skill catalog → Flow skills: includes · Major Maestro flow skill package.
- Skill catalog → BPMN skills: includes · Major Maestro BPMN skill package.
- Skill catalog → Case skills: includes · Major Maestro case skill package.
- Skill catalog → Agents skills: includes · Major agents skill package.
- Skill catalog → Coded apps: includes · Major coded apps skill package.
- Skill catalog → Platform ops: includes · Platform operations skill package.
- Skill catalog → Solutions: includes · Solution design and deployment skill package.
- Skill catalog → Troubleshoot: includes · Troubleshooting skill package.
- Skill catalog → Review: includes · Review and governance skill package.
- Agent layer → Claude/Cursor/Gemini: contains · Runtime-specific configuration directories and files.
- Claude/Cursor/Gemini → AGENTS/CLAUDE: bootstraps · Top-level guidance files feed agent initialization.
- Claude/Cursor/Gemini → Rules packs: defines · Commands and rules encode runtime behavior.
- Plugin layer → Agent layer: integrates with · Plugin metadata complements agent discovery.
- Validation hub → Task suites: organizes · Scenario suites are the main validation structure.
- Validation hub → Hooks/scripts: enforces · Local hooks and scripts catch issues early.
- Validation hub → GitHub Actions: runs in · CI workflows execute repository gates.
- Task suites → Skill catalog: exercises · Task fixtures validate skill instructions and outputs.
- Hooks/scripts → Skill catalog: checks · Hooks validate skill descriptions and installability.
- GitHub Actions → Skill catalog: verifies · Workflows validate, smoke-test, and lint skill content.
- RPA skills → UiPath services: targets · RPA workflows and activities operate against UiPath products.
- Flow skills → UiPath services: targets · Flow authoring and runtime interact with UiPath services.
- BPMN skills → UiPath services: targets · BPMN authoring and diagnosis reference UiPath platform services.
- Case skills → UiPath services: targets · Case workflows rely on UiPath runtime and service boundaries.
- Agents skills → UiPath services: targets · Agent lifecycle skills depend on UiPath agent and integration services.
- Coded apps → UiPath services: targets · Coded app guidance integrates with UiPath backends.
- Platform ops → UiPath services: targets · Platform operations directly cover UiPath services.
- Solutions → UiPath services: targets · Solution design and deployment span UiPath products.
- Troubleshoot → UiPath services: targets · Playbooks investigate failures in UiPath services and runtimes.