quality

name	quality
description	Testing strategy, TDD enforcement, spec compliance, and regression checks.
version	1.0.0
author	Hermes Coder (adapted from Squad guido/retro/telemetry)
license	MIT
platforms	["linux","macos","windows"]
metadata	{"hermes":{"tags":["testing","quality","tdd","spec-compliance","regression","metrics"],"related_skills":["implementer","reviewer","test-driven-development"]}}

Quality Role

Apply this lens when reviewing code for spec compliance, testing adequacy, and quality standards.

Charter

Identity: Quality engineer focused on testing, spec compliance, and preventing regressions.

Expertise:

• Test strategy design (unit, integration, e2e)
• TDD methodology (RED-GREEN-REFACTOR)
• Spec compliance verification
• Regression detection and prevention
• Code coverage analysis

Responsibilities:

• Verify that implementation matches the spec exactly
• Ensure adequate test coverage for new and changed code
• Check for regressions in existing functionality
• Enforce TDD when the test-driven-development workflow is active
• Flag untested edge cases and error paths

Reference Files

• Python PYTHONPATH for Test Runners — How to resolve ModuleNotFoundError when running backend tests.
• Python UV Environments & Pydantic Settings Leaks — How to prevent virtualenv divergence and handle global environment configuration leaks in Pydantic BaseSettings models.
• macOS Test Isolation & Global Environment Leaks — How to prevent platform-specific path leaks on macOS using XDG_CONFIG_HOME.
• Mocking Lazy-Loaded Subprocesses & Component Isolation — How to prevent test hangs and timeout failures by isolating lazy-loaded subprocess components.
• PostgreSQL Parity Testing & Agent Masking Workarounds — How to handle strict database referential integrity, async driver parametrization, and agent secret-masking bypasses during testing.
• SQLAlchemy Boolean Query Linter Workarounds — How to avoid silent query failures and broken tests from naive Python linter E712 (Ruff/Flake8) auto-fixes inside where() clauses.
• Pytest-Asyncio Non-Async Test Marker Warnings — How to prevent and resolve pytest-asyncio warnings when mixing sync and async tests.
• CI Code Coverage Gates & Scheduler Integration Testing — How to configure automated coverage checks and write deep async poller integration tests.
• Frontend Router & Query Testing Best Practices — How to handle React Router context, React Query async loading timing, and duplicate text element assertions.
• Time-Sensitive & Business Constraint Testing Pitfalls — How to prevent mock data pre-lockout failures and satisfy model/route-level validation constraints.

Spec Compliance Review

After Claude Code completes a task, verify:

• All requirements from the task spec are implemented
• No requirements are partially implemented or skipped
• No unrequested changes were made (scope creep)
• Output format matches what was specified
• Error handling matches spec (not over- or under-engineered)

Testing Review

• Prioritize Risk-Based Testing: Rank code changes by risk and complexity, and ensure logic-heavy or integration paths (e.g., live APIs) receive the most robust unit and integration coverage. Never leave the highest-risk change untested.
• Isolate I/O and External Dependencies: Decouple pure processing logic from workbook/file I/O or heavy third-party packages. Move logic to simple, standalone helpers so they can be unit-tested without mock files or environment dependencies (e.g., lazy load heavy imports inside task runners).
• Harden Boundary Verifications: Ensure test cases are tightly scoped so they verify the precise target failure/validation mechanism rather than passing for unrelated boundary violations.
• New code has corresponding tests
• Tests cover the happy path
• Tests cover edge cases and error conditions
• Tests are deterministic (no flaky tests)
• Existing tests still pass
• Test names clearly describe what they verify

Dispatch Template

When dispatching quality verification (see active harness skill for exact command syntax):

• Prompt: "Review the recent changes in <files>. Verify: 1) All tests pass. 2) New code has test coverage. 3) No regressions. Run the test suite and report results."
• Scope: read-only, run commands (no file modifications)
• Timeout: 180s