[gabrielseco]

Summary

Adds a Claude Code subagent that reviews the tests on a branch. It checks them against the repo's testing rules and proves each test fails when the behaviour it covers is removed.

Why

• A test that still passes with its fix reverted proves nothing. On fix(onboarding): refetch employment after basic info #1515 the first version of the test passed only because the mock already returned the new job title.
• Mocks encode assumptions about the API. Hand-built schemas, handlers that ignore the payload, and calling flow methods directly instead of clicking through the step all hid real bugs this week.
• A dry run on fix(onboarding): refetch employment after basic info #1515 found two gaps the author missed: a refetch-ordering claim that no test covered, and a PR description that said a failed update throws when it doesn't. Both are fixed there.

What changed

Toggle details

• .claude/agents/test-reviewer.md: the subagent. It reads CLAUDE.md fresh each run instead of keeping its own copy of the rules. It then:
  • checks each new or changed test against those rules, plus a short list of problems from recent PRs (mocks that hide the bug, inline schemas, skipped wiring, rendering before load, negative assertions with no wait, ordering claims, duplicated coverage);
  • removes the behaviour each test covers, runs the test, and restores the file from a backup. It never uses git checkout or git stash, so it can't lose uncommitted work;
  • reports findings with a table of mutation results. It doesn't edit, commit or push.
• CLAUDE.md: a new testing convention, "A test for a fix or feature must fail without it", which links to the agent.
• No library code changes.

Screenshots

N/A

Related Resources

• Dry run: fix(onboarding): refetch employment after basic info #1515.

Testing

• Dry run on fix(onboarding): refetch employment after basic info #1515. It reported 3 Medium and 2 Low findings, plus a mutation table. Afterwards git status matched the starting state.
• Feature flag: N/A

🤖 Generated with Claude Code

---

Note

Low Risk
Documentation and Claude agent config only; no product code, APIs, or CI behaviour changes.

Overview
Adds a Claude Code test-reviewer subagent (.claude/agents/test-reviewer.md) for branch test review: it reloads testing rules from CLAUDE.md, audits new/changed tests (including repo-specific pitfalls like MSW mocks that hide bugs, inline schemas, and ordering claims without held responses), mutates production code or waits to confirm each test fails when the covered behaviour is removed, then reports High/Medium/Low findings without editing tests permanently.

CLAUDE.md gains a convention that fixes/features must have tests that fail without the change, with a link to the subagent. No application or test runtime code changes.

^{Reviewed by Cursor Bugbot for commit 7afcd28. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Deploy preview for adp-cost-calculator ready!

Project:	adp-cost-calculator
Status:	✅  Deploy successful!
Preview URL:	https://adp-cost-calculator-fznhccyvf-remotecom.vercel.app
Latest Commit:	7afcd28

Deployed with vercel-action

[github-actions]

Deploy preview for remote-flows ready!

Project:	remote-flows
Status:	✅  Deploy successful!
Preview URL:	https://remote-flows-gtbev8ge0-remotecom.vercel.app
Latest Commit:	7afcd28

Deployed with vercel-action

[github-actions]

📊 Coverage Report

✅ Coverage increased! 🎉

Metric	Current	Previous	Change	Status
Lines	89.58%	89.53%	+0.05%	🟢
Statements	89.45%	89.40%	+0.05%	🟢
Functions	88.47%	88.41%	+0.06%	🟢
Branches	81.20%	81.14%	+0.06%	🟢

Detailed Breakdown

Lines Coverage

• Covered: 5389 / 6016
• Coverage: 89.58%
• Change: +0.05% (4 lines)

Statements Coverage

• Covered: 5454 / 6097
• Coverage: 89.45%
• Change: +0.05% (4 statements)

Functions Coverage

• Covered: 1427 / 1613
• Coverage: 88.47%
• Change: +0.06% (1 functions)

Branches Coverage

• Covered: 3343 / 4117
• Coverage: 81.20%
• Change: +0.06% (4 branches)

✅ Coverage check passed

[cursor]

Stale Bugbot comment from a previous run.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 7d1d4af. Configure here.}

[gabrielseco]

Closing: keeping this subagent and the testing rule local instead of shipping them in the repo.