Agents In Use
What’s new todayLast update
18 additions

Watch Codex repair a Datasette plugin compatibility failure

Willison recorded upgrading datasette-checkbox for a changed permissions API. He first repaired a test that missed a real page failure, then gave Codex the failing runner and upgrade documentation.

Added simonwillison.netAgent: Codex
TRY A BOUNDED VERSION

Prompt to try

Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.

Run [test command] against [new version] in a sandbox. Confirm the test catches the real compatibility failure. Read [upgrade guide], make the smallest fix and rerun the suite. Report changed behavior and test results; stop before release.

What happened

Willison documents the plugin upgrade and the verification process in a recorded walkthrough.

Reported by Simon Willison

How the agent did it

  1. Run the plugin with the newer Datasette version and reproduce the page error.
  2. Strengthen the test so it catches the missing permission method.
  3. Have Codex read the upgrade notes, apply the compatibility fix and rerun tests.

Where a person stepped in

Willison exposed the real failure and improved the test before delegation. His bypass-approval command is not recommended by this entry.

A person approved each step before the agent continued.

Open original on YouTube ↗ · If the embed is unavailable, view the original post.

More details

About this account

Codex iterated against a failing command and migration documentation in a non-interactive run.

Companion article fetched; YouTube oEmbed confirms the recording and author. Exact duration not verified, so omitted.

Credit: Simon Willison

Source published 6 Nov 2025

Source checked 14 Oct 2026

Tools and topics

Codex CLI · Datasette · uv · pytest · sqlite-utils

Code · Testing

Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.

Video notes

@swillison · YouTube

  • A passing test can miss a real page error; reproduce the behavior first.
  • Give the agent a failing runner and the precise migration guide.

Related use cases