# Executor Agent: Exactly the Task Given, Reported With Evidence > The Executor Agent is the AI coding task executor inside a software build in the Ai1 platform by MyZone AI: a build orchestrator hands it one task at a time, and it reads every guideline the task names, does exactly that task and reports back with evidence. Canonical page: https://myzone.ai/pages/agents/ai-executor-agent Part of Ai1, by MyZone AI. Book an Executor Agent walkthrough: https://calendly.com/d/ct6h-tcy-8qf/ai1-demo?a1=Executor%20Agent&utm_source=myzone.ai&utm_medium=agent-page&utm_content=ai-executor-agent-final Last updated: 2026-10-05. Reviewed by the MyZone AI team. One build task, done exactly as briefed and reported with evidence. ## At a glance - **Who hands it work:** A build orchestrator, one task at a time; you do not brief it directly - **Before starting:** Reads every skill, document, guideline and file the task points to - **Unclear brief:** Asks the orchestrator instead of guessing - **The work:** Exactly what was asked, nothing more, nothing less - **Every report:** What was done, the evidence and anything blocking it - **What it never does:** Merge, release or deploy, or carry one project into another ## What the Executor Agent does - **Loads the context it is given:** Before any work, it reads every skill, project document, development guideline and file the task points to, and follows them exactly. - **Asks instead of guessing:** If a brief is incomplete, unclear or contradicts itself, it asks the build's orchestrator before doing anything. - **Does exactly what was asked:** It carries out the task to a high standard, following the project's development guidelines. Nothing more, nothing less. - **Reports back with evidence:** It returns a short report of what it did, the evidence, and anything that blocked it or needs a decision. - **Saves code only when told:** It commits and pushes code work only when the task says to. - **Keeps the task board current:** When a task asks for it, it updates the build's task board. ## How the Executor Agent differs from a general coding assistant A general coding assistant can drift beyond the task, fill gaps with guesses and carry one project into another. The Executor Agent starts each task knowing nothing, reads only what the task points to and does exactly that task. When a brief is unclear it asks the orchestrator, it commits code only when told, and every task ends with a short report and its evidence. ## How it works 1. **Handed a task** (When the orchestrator has a task): Your build's orchestration hands it one task. 2. **Read** (Before any work): It reads the brief, the guidelines and the files the task points to. 3. **Ask if needed** (Only when the brief is unclear): Anything missing or unclear goes back to the orchestrator as a question. 4. **Do the task** (Once the brief is clear): It does exactly what was asked, and no more. 5. **Report** (As soon as the task ends): A short report with evidence goes back to the orchestrator. ## When to use it - **A homepage needs a new section built to the project's guidelines:** It reads the guidelines, makes the change as specified and sends back a short report with evidence. - **A brief does not say which form to update:** It stops and asks the orchestrator which form before doing any work. - **A follow up task arrives on the same project:** It reuses what it has already read and reloads only what changed. - **Software build tasks that must follow development guidelines:** It reads every guideline the task names before starting, then does the task as written. ## What you get - The task carried out exactly as briefed - A short task evidence report: what was done, the proof and any blockers - Questions to the orchestrator whenever a brief is unclear - Code commits and pushes, only when the task asks for them - Task board updates, when the task asks for them ## Example: build task report Example with a fictional company. Names, people and figures are invented to show the agent's output. Any resemblance to a real company or person is unintended. Industry facts are real and cited with their sources. What it was asked: Add a "Book a service visit" form to the Services page that matches the design guide, checks required fields, works on a 390 px phone screen, meets WCAG 2.2 AA and comes with tests; commit to the task branch but do not merge. The report follows the Executor Agent's five steps from its public guide: handed a task, read, ask if needed, do, report. All figures, times and the company are invented for the sample; no real website was built or changed. The only real fact is the WCAG 2.2 minimum target size, checked on the W3C page. Run date: 1 October 2026. ### What it found - The brief did not say whether a phone number was required, so the Executor Agent paused and asked; the build orchestrator checked with the client's project lead, who chose optional. - The booking form was built exactly to the brief: 4 files changed, all 5 "done when" lines met, 12 of 12 tests passing, and every form tap target 44 by 44 pixels. - An existing 20 by 20 pixel "Call us" icon may fall below the WCAG 2.2 minimum of 24 by 24 pixels; it was outside the task, so it was flagged, not changed. ### Sources - [Understanding Success Criterion 2.5.8: Target Size (Minimum) | WAI | W3C](https://www.w3.org/WAI/WCAG22/Understanding/target-size-minimum.html), World Wide Web Consortium (W3C), Web Accessibility Initiative (accessed 2026-10-01) ## Guardrails - It is not a stand alone assistant: you do not brief it directly, and it only runs tasks handed to it by your build's orchestration. - It never widens a task, and asks rather than guesses when a brief is wrong or incomplete. - It commits and pushes code only when the task says so; merges, releases and deployments belong to the DevOps Agent. - Project knowledge arrives with each task and is never carried into another engagement. ## Frequently asked questions ### Why hand an AI agent one well-defined software build task at a time? A small task with a clear brief is easier to check. In Ai1 by MyZone AI, a build orchestrator decides what to build and splits the work, and the Executor Agent carries out one task at a time exactly as briefed, so scope does not quietly grow. It starts each task with no project knowledge of its own, reads only what the task points to, does exactly that task and reports back. ### What should a task evidence report contain? What was done, the proof that it was done and anything that blocked the work or needs a decision. The Executor Agent ends every task with a short report like this, sent back to the orchestrator that handed it the task, so you know the task was really done. ### How do you make an AI coding agent follow development guidelines? Name the guidelines in the task and have the agent read them before it starts. The Executor Agent reads every skill, project document, guideline and file a task points to and follows them exactly. If the brief is unclear, wrong or contradicts itself, it asks the orchestrator rather than guessing, and never goes beyond the task it was given. Project knowledge arrives with each task and is never carried into another engagement. ### Is the Executor Agent available, and can I hand it work myself? Yes, it is ready to use. You do not brief it directly: it runs tasks handed over by a build's orchestration inside a software build, one task at a time, and reports back with evidence for each one. ## More on this topic - [Multi-Agent Development: Why the Future of Software Is Built by AI Teams](https://myzone.ai/pages/blog/multi-agent-development): Discover how multi-agent AI systems are transforming software development. Teams of specialised AI agents build products 10x faster at lower cost. ## About Ai1 Ai1 is the AI operations platform by MyZone AI, where each client runs on its own private server. The Executor Agent is included on every Ai1 level, including Developer Core, with no per-agent charge. It works alongside the other Ai1 agents on your account. Pricing: https://myzone.ai/pages/services/ai1-pricing. Security: https://myzone.ai/pages/security.