Artifacts in Antigravity: reviewing AI work with evidence

November 3, 2026

Updated:

Guide
Artifacts in Antigravity: reviewing AI work with evidence

What are Artifacts in Antigravity?

An artifact is a structured output that the agent in Google Antigravity creates to work on and show me what it's thinking, what it has done: implementation plan, code diff, architecture diagram, screenshot, browser recording, and walkthrough. According to the official documentation (read 03/11/2026), artifacts enable asynchronous collaboration: I don't need to track every tool call, just review the outputs at key checkpoints. This post doesn't promise that artifacts eliminate AI mistakes: artifacts are evidence for the reviewer to check, and how to check them is still my responsibility. The post goes through each artifact type, how to comment, review policies, the Review frame, and /diff. For a general introduction see What is Antigravity?.

Diagram of artifacts in Antigravity: implementation plan, comments or Proceed, diff, screenshot, browser recording, walkthrough, Review frame and line comments back to the agent
Review loop using artifacts: plan, evidence, comments, revisions, then keep changes. Diagram: OPCAIOS.

What kinds of artifacts does Antigravity create?

Artifacts in Antigravity fall into five main groups, according to the documentation pages on artifacts, implementation-plan, walkthrough, screenshots, and browser-recordings (read 03/11/2026). Most artifacts are created in the agent's Planning mode.

Artifact When it appears What to review
Implementation plan Before editing code, or when calling /plan Approach, scope, verification steps
Code diff When the agent writes a file Each line changed
Screenshot When the agent's browser child needs me to see page state Interface is correct or not
Browser recording Each time the agent's browser child acts What the agent clicked, typed
Walkthrough When work is done Summary of what changed

About task list: the current documentation has no separate page (read 03/11/2026). The artifact-review page only mentions Planning mode dividing work into "task groups", and the changelog mentions "task list" as one artifact type.

What is an implementation plan and how do I review it?

An implementation plan is an artifact where the Antigravity agent designs changes to the code before making them, containing technical details to fix and written for me to review. According to the Implementation plan page (read 03/11/2026), unless the review policy is set to Always proceed, the agent usually stops and asks me to review the plan before editing files. I have three choices:

  1. Proceed: click in the conversation or at the top of the artifact for the agent to follow the plan.
  2. Comment then Proceed: leave notes for the agent but let it continue.
  3. Review: click Review at the top of the artifact to see all comments again and send them as one feedback message instead of proceeding right away.

After I Proceed or send Review, the agent either revises the plan and asks for review again, or starts implementing. The /plan command forces the agent to survey the code without writing files, ask back if the request is unclear, then write the plan. The /plan page recommends it for large refactors, unclear requests, or database migration. This is like Plan mode in Claude Code.

How does commenting on an artifact for AI to fix work?

Commenting on an artifact is how I point the Antigravity agent to the right place that's wrong, instead of writing a long message describing it all over again. According to the Mentions and quoting page (read 03/11/2026), when I select a passage in an artifact like an implementation plan, a small window shows two buttons: Comment to leave a review note, Quote (Cmd+L on macOS, Ctrl+L on Windows and Linux) to bring that passage into the next message. Screenshots are also saved as image artifacts and I can comment on them too. Four tips:

  1. Each comment says one thing: "skip this step", "use the existing library", not many ideas mixed.
  2. Write the reason short so the agent doesn't repeat the mistake in another step.
  3. Name the acceptance criteria: "page must load when not logged in".
  4. Read the revised plan again, don't assume the agent did it right.

Example scenario, not a real project: the plan proposes adding a scheduling page and suggests installing a new calendar library. The reviewer comments right on that step: "use the browser's built-in date picker, don't add a library", then clicks Review.

What artifact review policy should you set?

The artifact review policy decides whether the Antigravity agent stops and waits for you before following the plan or not. According to the Artifact review page (read 03/11/2026), when you open a new conversation you choose between two modes: Planning (plan carefully, create artifacts) and Fast (do it directly, suitable for small tasks like renaming a variable or running a command). In Planning mode, the review policy is in the Agent tab of Settings, with two options:

  1. Request review (recommended): the agent always stops and waits for you to review when creating a plan or diff.
  2. Always proceed: the agent never stops to wait for review, it continues on its own.

The review policy is different from the command execution permission mode. The /goal page makes clear that goal mode only changes how the agent stops, it does not change tool permissions or artifact review policy. Newcomers should use Request review; Always proceed is only suitable for trial projects that have Git and backups. Details about permission levels like Default, Auto or Turbo are in the post safety modes in Antigravity.

What do walkthroughs, screenshots and browser recordings prove?

Walkthroughs, screenshots and browser recordings are three types of artifacts that prove what the Antigravity agent has done, different from what the agent tells you itself. According to documentation read on 03/11/2026, a walkthrough is created when the agent finishes work, summarizing the changes so you can grasp the code state even if you didn't follow each step. For browser work, a walkthrough usually comes with screenshots and browser recordings. Screenshots are taken by the browser agent of the page or part of the page when needed for you to see. Browser recordings capture the browser agent's actions, you can replay them in a loop; the feature page lists the format as WebM video. Three limits to remember:

  1. A walkthrough is a summary written by the agent, so you still need to check it against the actual diff.
  2. A screenshot only shows one moment in time, it doesn't prove the page works correctly in every scenario.
  3. A recording shows the agent has acted, it doesn't yet prove the result is correct for the business requirement.

Treat artifacts as evidence to ask questions with, not as a final confirmation. For how the browser agent works see browser agent of Antigravity.

How do you use Review changes on the app and /diff on CLI?

Review changes is Antigravity's diff viewing frame to check each file the agent has edited before keeping it, sending line comments and committing to Git. According to the Review changes page (read 03/11/2026), on the app you open the Review frame from the Overview panel or from the "N files changed" tab below a response. The frame has three viewing modes:

Viewing mode Shows
Uncommitted Staged and unstaged changes (default)
Branch All changes on the current branch compared to the base branch
Agent Edits Files the agent edited in this conversation

Line comments are attached to the next message, the agent receives them with the file name and line number. Important note: the Discard unstaged changes button takes effect immediately without asking again and deletes untracked files from disk. On CLI, the /diff command opens a full-screen diff viewer with three modes All Changes, Per Turn and Commit Tree; press C to comment, exiting will ask whether to send unsent comments or not.

What are the steps in the process of reviewing AI work using evidence?

The evidence-based review process is a habit of only accepting the agent's changes after you have seen the plan, diff and verification results, rather than trusting its summary. OPCAIOS's suggestion, assembled from features in the Antigravity documentation:

  1. High-risk work: use /plan or Planning, Request review policy.
  2. Comment on the exact step to change in the plan, press Review.
  3. Write the criteria that you can verify when done: test commands must pass, the page must show what.
  4. Work finished: open the Review frame on Agent Edits, read each file.
  5. Interface: view screenshots, recordings, then open the page yourself to check.
  6. Read the walkthrough last, compare it with the diff, then commit.

For people who don't write code, these six steps are the minimum safety layer; see more in safe process for people who don't know code.

OPCAIOS's Experience

In OPCAIOS's setup, Antigravity is the coordinating agent on Mac, running Antigravity app version 2.28.x, Antigravity IDE version 2.5.x, and CLI agy version 1.4.12, used in parallel with Claude Code and Codex. Antigravity has handled some website management features and menu fixes for us. Here's what we observed on our machine, checked on 03/11/2026:

  1. The agent uses a separate Chrome profile, isolated from the personal browser.
  2. The browser recording directory is currently empty, so we have no recordings to showcase and cannot confirm the level of sub-agent browser use.
  3. Website code is only edited and deployed from a single machine. Reason: we once had over 1,500 files showing as different just because of line ending differences when deploying from Windows. Diffing thousands of files can't be reviewed by eye, so we had to clean up the root cause first. We haven't verified whether the Review frame's whitespace hiding option can also hide line ending differences.

We haven't measured review time or the error-catch rate from artifacts, so we have no numbers to share.

Frequently Asked Questions

Where are artifacts in Antigravity saved?

According to the Agent settings page (checked 03/11/2026), the app's local data folder contains artifacts, knowledge items, and other data, and by default the agent can only access the project folder and this data folder. In the app, you view artifacts in the sidebar and review pane; in CLI, open the artifact list with Ctrl+R. Don't push this folder to a public Git repo since it may contain private content.

Can I turn off the agent waiting for plan review?

Yes. In the Agent tab of Settings, change the artifact review policy from Request review to Always proceed, and the agent won't wait for review anymore. Fast mode for small tasks also skips the planning step. The documentation recommends Request review. Only disable the review step for experimental projects that have Git and backups.

How is commenting on artifacts different from sending a regular message?

Comments attach to a specific passage or line of code, so the agent knows exactly what to fix. With diffs in the Review pane, the agent gets the file name and line number. With plans, you can gather multiple comments and click Review to send them at once. Messages work for big direction changes; comments work for detailed fixes.

Is a walkthrough enough to trust the agent did the work right?

No, it's not enough. A walkthrough is a summary the agent writes after finishing. It can be incomplete or overly optimistic. Compare it against the real diff in Agent Edits mode, rerun tests, and open the page yourself to see. Screenshots and browser recordings are supporting evidence, not a substitute for checking yourself.

How should a newcomer start using artifacts?

Start with Planning mode and Request review policy for anything that touches real code. Each time the agent presents a plan, read it fully and leave at least one comment if you spot something unclear. When done, open the Review pane and read each file before committing. If you want to run multiple agents, read more about Agent Manager in Antigravity. Want to learn this workflow with a guide, check out OPCAIOS's AI application training course.

Source: Google Antigravity documentation at antigravity.google/docs: artifacts, artifact-review, implementation-plan, walkthrough, screenshots, ide/browser-recordings, review-changes, ide/review-changes-editor, cli/artifacts, plan, commands/goal, at-mentions, features, agent-settings, changelog (checked 03/11/2026). Not verified: artifact task list details (no dedicated page), whether whitespace hiding can also hide line ending differences. Experience is from OPCAIOS.

Translated from Vietnamese with AI. Read the original.

Want to apply this in your business?

Book a free 30-minute call (Vietnam time, in English or Vietnamese).

Book a call