Documentation navigation
Getting started Quickstart Installation Core concepts Product tour
Building trees Creating nodes Prompt blocks Connectors Roles and governance Skills Co-pilot Collaboration
Run and inspect Execution Runs and traces Versions and rollback
SDK Python SDK Node.js SDK
Integrate API MCP and bundles
Providers AI connections Hosted providers for testing
Deploy Self-hosting Orchetree Cloud Environment variables

Skills

A skill is a reusable instruction asset stored at project level and attached to one or more nodes.

Workflow

  1. Find or create a skill for the project.
  2. Submit it to the mandatory security review.
  3. Store it in the project registry only after a passing verdict.
  4. Attach the reviewed skill to a node.
  5. Run the node and inspect the trace.

The security review checks for unsafe instruction patterns, prompt injection, and possible data-exfiltration behaviour. A failed review blocks storage and use through every path, including the UI, API, SDK synchronization, and Co-pilot actions.

Runtime behaviour

Passing skills attached to a node are loaded into its runtime instructions. Nodes can disable runtime skill injection, and a skill can be marked required so execution fails clearly if it cannot be loaded.

Project-level registration avoids repeating the same review whenever an approved skill is reused by another node.

Operational guidance

  • Treat third-party skills as code and review their source and license.
  • Keep credentials and private data out of skill instructions.
  • Re-run tests when a skill changes a node’s available tools or expected output.
  • Inspect the node trace to confirm which skills were loaded.

See Nodes, Runs and traces, and Roles and governance.