agent-builder-skill

To safely perform mid-execution canvas mutation, you must inspect existing nodes and wiring before adding tools or enabling skills. This prevents duplication and applies changes via a single entrypoint, ensuring newly added capabilities appear only on the next turn through soft-reload semantics.

Canvas inspection in agent orchestration is the process of reading existing tools, skills, and teammates wired to an agent. It is needed before any mid-execution mutation to avoid duplicating workflow components and to coordinate changes safely across delegates.

To delegate subtasks without duplicating teammates, inspect the canvas to check existing delegates, then spawn a team-lead delegate subagent. This controlled growth ensures new specialist agents are added safely without overlapping current workflow wiring.

Yes, you can persist a new empty workflow and enable a memory skill during runtime through mid-execution canvas mutation. The agent dispatches these mutations to a single entrypoint tool, applying the changes safely with soft-reload availability on the next turn.

Inspecting the canvas before mutating the workflow is a strict rule to prevent duplication and confusion. It ensures the agent reads existing nodes, edges, and teammates before adding tools or skills, coordinating changes safely across the Master Skill and delegates.

The primary limitation of mid-execution workflow mutation is that newly added capabilities do not appear immediately. Changes are bound to a single entrypoint tool and require soft-reload availability, meaning new tools and skills only activate on the next turn.