Willison ships shot-scraper 1.10 with agent-ready video demos

Simon Willison shipped a small but pointed addition to his browser-automation utility this morning, aimed squarely at coding agents that want to prove their work actually runs. The new subcommand reads a YAML storyboard, drives a browser through the listed steps with Playwright, and saves the result as a video file an agent can drop into a pull request or release note.

According to Willison's release post, "shot-scraper video is a new command introduced in today's shot-scraper 1.10 release which accepts a storyboard.yml file defining a routine to run against a web application and uses Playwright to record a video of that routine." The command outputs WebM or MP4, supports authentication via JSON cookie files, and uses Pydantic to validate the storyboard schema. The worked example records Datasette's new CSV, TSV and JSON bulk insert feature.

The framing that makes this worth paying attention to is the agent ergonomics, not the video format. Willison argues that the command's --help output is detailed enough that a coding agent can use the tool with no separate prompt, calling it "kind of like bundling a SKILL.md file directly inside the tool." He used GPT-5.5 xhigh inside Codex Desktop to generate the example storyboard and the feature documentation, and is candid that "this is a great example of the kind of feature that I almost certainly wouldn't have taken on without coding agent support."

The honest caveat is that this is a single maintainer's release of a small utility, not a benchmark. The original issue had been open since February 2024, and only landed once an upstream change in playwright-python 1.61.0 fixed the recording behaviour, so the agent did not magic away the hard parts so much as finally have something stable to build on. What the post does not address is how often the agent-generated storyboards work without human cleanup, or whether reviewers will trust a self-produced demo video over running the change themselves.

For solo maintainers and small teams who would never sit down to screen-record every feature, the move worth copying is letting the agent produce a walkthrough as part of the PR. Whether that turns into a release-notes accelerant or just more clip clutter depends on how review culture adapts.