A QA engineer agent reviews pull requests for test coverage gaps, writes missing tests, and runs acceptance checks against the test environment before a release is approved. It is a distinct role from a PR reviewer or tech lead: where the reviewer checks logic and style, the QA agent focuses on correctness, edge cases, and regression risk.
A role-specific prompt encodes your team's quality bar: which test types are mandatory, what coverage thresholds trigger a block, and how to categorize bugs by severity. A generic agent will write tests but may not enforce the policies that actually protect your release pipeline.
What this agent owns
- Review PRs for missing unit, integration, and end-to-end test coverage
- Write test cases for edge cases and error paths not covered by the author
- Run the test suite against staging and report failures with reproduction steps
- Categorize defects by severity and link them to the triggering change
- Maintain test utilities, fixtures, and shared test helpers
Recommended model: Claude Sonnet
Test writing and coverage analysis are pattern-matching tasks where Sonnet is fast and accurate without requiring deep reasoning.
Example tasks
- Write missing integration tests for a new payments flow
- Identify untested error paths in an authentication module
- Generate a regression test suite for a recently fixed bug
- Audit a PR's test coverage and report gaps before it merges
# create an agent from this template, then start it
$ fleet agent create --name qa-engineer--vendor claude-code --template <template-name>
$ fleet agent start qa-engineer
Find the exact template name with fleet template list.
