test-checklistlisted
Install: claude install-skill DazzleML/dazzle-claude-code-config
# Test Checklist Skill
Produce a **human-runnable test checklist** alongside significant automated test work. This document complements automated tests — it captures what the mocks can't, and it serves as a durable record of "what should we test" for each phase/feature.
## When to use
Trigger this skill broadly — **err on the side of producing a checklist**. Automated tests catch regressions in mocked paths; human checklists catch the subtle behavior drift that mocks miss. The cost of producing one is low; the cost of shipping a "pure refactor" that silently breaks metadata preservation or subprocess handling is high.
Produce a checklist when you finish writing automated tests for:
- A new Phase, epic, or major feature
- A new CLI command or user-visible surface area
- A refactor that touches dispatch, configuration, file I/O, subprocess execution, metadata handling, or discovery — **even if the refactor claims no user-visible change**
- Any commit with a substantive CHANGELOG entry (Added, Changed, or Fixed for a real bug with user impact)
- Anything where a subtle regression could affect users in non-obvious ways
If the user explicitly says "let's write tests" or "add tests for this," ask at the end: **"Do you also want a human test checklist for the phase/feature we just tested?"** Default to yes unless the work is clearly trivial.
## When NOT to use
Skip the checklist ONLY for genuinely trivial work:
- Single-line bug fixes with truly no interaction surface
- Doc