How to use Claude to add tests with subagents
Claude Code split a testing backlog into non-overlapping tasks, launched 26 subagents, and helped the author review the resulting tests over several days.
news.ycombinator.com
Prompt to try
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
For [repository], develop a technical plan for adding tests toward [coverage target]. Divide the work into non-overlapping sections and delegate each section to subagents. Report changes and support a careful review of whether each test checks meaningful behavior.
Reach people exploring AI agents
See placements →What happened
The author reported about 20 minutes for the generated work and judged the reviewed tests comparable to ones they would have written.
Reported by esperent on Hacker NewsHow the agent did it
- Specify the testing work and develop a technical plan.
- Ask Claude to divide the work into non-overlapping pieces.
- Launch subagents to implement tests for those pieces.
Where a person stepped in
The author spent a couple of days reviewing the tests and asked Claude to review sections carefully.
A person set the goal and reviewed the result.
An author adding tests to an old project began by carefully specifying the work. Their target was 70% coverage, which they said required a few thousand unit tests. They developed a technical plan with Claude Code before deciding that the work could be divided among subagents.
The author asked Claude to separate the project into non-overlapping pieces and send as many agents as it could to work on them. Instead of the three or four agents they expected, Claude launched 26. The author reported that the tedious work finished in about 20 minutes, compared with their optimistic estimate of several months to do it themselves.
That generation time was followed by a couple of days of human review. The author checked whether the tests were doing their job properly and also asked Claude to review separate sections carefully. They expressed confidence that the resulting tests were as good as ones they would have written. The post describes a coverage target, but does not state a measured final coverage percentage.
Who it's for: Developers facing a large testing backlog with work that can be divided into independent sections.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · esperent
Frequently asked questions
Was the entire task finished in 20 minutes?
The author reported about 20 minutes for the subagent work, followed by a couple of days reviewing the tests.
Did the post confirm 70% coverage?
No. It says the author was aiming for 70% coverage, but does not report a measured final percentage.
More details
Remaining steps
- Review the tests over several days and ask Claude to inspect sections.
About this account
Claude launched 26 subagents to work on separate portions of the testing task.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=46749612. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: esperent on Hacker News
Source published 25 Jan 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · Subagents
Coding agents
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.