v1-fix-tests

name	v1-fix-tests
description	Use when a failing test suite needs systematic repair. Triggers on "fix tests", "tests failing", or "make tests pass".
allowed-tools	["Bash","Read","Edit","Grep"]

Fix Failing Tests

Use this when you have failing test output. The goal is to fix ALL failing tests, re-run tests, and iterate until everything passes.

Usage

Typical invocations:

• Claude Code: /v1-fix-tests
• Codex: invoke v1-fix-tests from the skills menu or use $v1-fix-tests

Use this after you've pasted test failure output into the conversation.

What It Does

1. Understands ALL Failures

   • Reads entire test log (pytest / Jest / parallel tests)
   • Checks "Test Results" summary for ALL failing tests (looks for ✗ or FAILED markers)
   • Checks "Failed Test Details" for MULTIPLE failing test groups
   • Notes ALL test groups that failed (backend, frontend, linting, etc.)
   • Creates list of all failures before starting fixes
   • Confirms the failing command is a reliable feedback loop for the user-visible problem; if not, first narrow or rebuild the loop using the v1-debug skill

2. Fixes EACH Failure

   • For EACH failing service/group:
     • Opens failing test files and implementation files
     • Makes smallest, clearest change
     • Prefers fixing implementation over changing tests
   • Fixes ALL failures before proceeding

3. Re-runs Tests for EACH Fix

   • Backend: pytest (or with -k for specific tests)
   • Frontend: npm run lint or npm run test
   • Verifies each group passes before moving to next
   • Repeats if tests still fail

4. Re-runs Full Test Suite

   • CRITICAL: Always re-runs original command used to run tests
   • Catches hidden failures or new failures from fixes
   • Only stops when full suite passes with zero failures

Important Notes

• Don't stop after fixing one failure - check for multiple failing groups
• Always re-run full test suite after fixes
• Parse entire error output - summary AND detailed sections
• Prefer fixing code over weakening assertions; only change tests when the test is wrong or the intended behavior changed. When a test itself must change, follow the mock-discipline and assertion rules in v1-write-tests rather than restating them here.
• If the failure is flaky rather than a real regression, hand off to v1-debug to stabilize the reproduction (measure the failure rate; isolate time, randomness, filesystem, and network) before patching.

Testing Commands

Backend (pytest):

pytest                              # All unit tests
pytest -k "pattern"                 # Specific tests
pytest tests/integration/           # Integration tests

Frontend:

npm run lint
npm run test