What problem does it solve? Delegating development work to a remote AI agent requires understanding Netlify's agent task model: tasks run against the pushed repository (not local changes), are asynchronous with no completion callbacks, and output lands on a new branch with a Deploy Preview. This Skill encodes those semantics so tasks are created, polled, and reviewed correctly without misusing undocumented APIs. ## Core Features & Use Cases - Remote task creation: Launch Claude, Codex, or Gemini agents via netlify agents:create with prompt, branch, model, and project targeting options. - Task lifecycle management: List, inspect, and stop agent tasks with netlify agents:list, agents:show, and agents:stop, polling until a terminal status (done, error, cancelled). - Safe delegation guardrails: Enforces documented CLI surfaces only, requires explicit user approval before billable runs, and prevents delegating work that depends on unpushed local changes. - Use Case: Get a second opinion from a different model on your implementation by delegating a code review prompt to a remote agent, then polling for its result on a new Deploy Preview branch. ## Quick Start Ask the assistant to run a Netlify agent task, for example: create a remote agent task with the codex agent to audit this site's codebase for accessibility issues, then poll the task until it completes.
