mcpField
P
parallel-test-fixing

parallel-test-fixing

github.com/revitalb-portTesting & quality054 installsUpdated 10w ago
Repository

About parallel-test-fixing

A testing skill for developers who need to repair a failing test suite. It groups failures by test file, assigns independent work to parallel subagents, and verifies the fixes with a final full-suite run.

Key features

  1. 01Run and parse the full test suite
  2. 02Group failures by test file
  3. 03Delegate fixes to parallel subagents
  4. 04Collect changes and flag conflicts
  5. 05Rerun tests to verify fixes

Use cases

  1. 01Divide many failing tests across agents
  2. 02Avoid edit conflicts between test fixes
  3. 03Check whether repairs restore the full suite

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/revitalb-port/skills-registry-demo --skill parallel-test-fixing

After installing, type/parallel-test-fixingin Claude Code to confirm it loaded.

Details

From the project README

Parallel Test Fixing

Speed up fixing a broken test suite by distributing failing tests across parallel subagents.

Workflow

1. Run the Full Test Suite
npm test -- --no-coverage 2>&1 || true

Capture the output and extract all failing test files.

2. Group Failures

Parse the test output for failing files:

  • Jest: FAIL src/components/Button.test.tsx
  • Vitest: FAIL src/utils/format.test.ts
  • Pytest: FAILED tests/test_api.py::test_create_user

Group by file — each file becomes one task.

3. Launch Parallel Subagents

For each failing test file, launch a generalPurpose subagent:

Task: Fix the failing tests in <file>

The test file is: <path>
The test command is: <command to run just this file>
The error output was:
<paste the relevant failure output>

Steps:
1. Read the test file and the source file it tests
2. Understand why each test is failing
3. Fix the source code (preferred) or update the test if the test is wrong
4. Run the single test file to confirm it passes
5. Report what you changed and why

Launch all subagents simultaneously — they work in parallel since each touches different files.

4. Collect Results

As each subagent completes, collect:

  • Which tests were fixed
  • What files were changed
  • Whether the fix might conflict with another subagent's changes
5. Verify

Run the full test suite one more time to confirm everything passes:

npm test

If there are new failures (from conflicting fixes), resolve them sequentially.

Tips

  • If two failing tests share the same source file, assign them to the same subagent to avoid edit conflicts
  • Set a timeout — if a subagent is stuck for 5+ minutes, check its progress
  • For large test suites (50+ failures), batch into groups of 5-10 per subagent rather than one-per-file
  • Use best-of-n-runner subagents if you want isolated worktrees for each fix attempt

Comments

Markdown supported