Tests That Catch Bugs
Get tests that fail when the code is wrong — and proof that they do.
- For
- Developers who don't trust their test suite
- Use it when
- When your tests are green but you're not sure they would catch a real bug.
- Trust
- Made by Launch7 · not yet testedBuilt by Launch7 and tried end to end in a general chat assistant on 2026-08-25 while it was made. It has not yet been through Launch7's full tests.
Try it like this
Write tests for this function that would actually catch a bug.
This works once your AI has the helper’s instructions. Set that up first — about a minute.
What you’ll get
- Tests whose expected values come from the specification, not from running the code
- A check that each test fails against a deliberately broken version
- Failing tests reported as findings about your code
- An answer to 'what bug would this suite most likely miss?'
What you need
- You give it
- The code, and what it is supposed to do.
- Works with
- Not available yet
Use it with your AI
Built — the download isn’t published yet
Launch7 has built this, but hasn’t published it for download yet. It will appear here, with steps for your AI, as soon as it is.
Available today from the community — not made or tested by Launch7:
- qa-test-planner — davila7/claude-code-templates, MIT
Before you rely on it
- It needs to know what the code should do — it won't guess the intent.
- Setting up the broken versions takes longer than writing plain tests.
Why it’s better than just asking
Each test is run against a deliberately broken version of the code and must fail; tests that stay green are removed.
What next
Coming: the Ship Safely Kit — plan, test and review every change before it ships.
Technical details — for the curious and for developers
Phrases that should switch it on (as an installed skill)
“write tests for this” · “are my tests any good”
Where it comes from
A pack built by Launch7.
| Source | Role | Repository | Licence |
|---|---|---|---|
| qa-test-planner | merged | davila7/claude-code-templates | MIT |
Test record
| Date | Where | Kind | Result | Notes |
|---|---|---|---|---|
| 2026-08-25 | a general chat assistant | Run while building | pass (3 cases) | Run end to end in a general chat assistant while the pack was built. |
Publication checks
Safe and sourced: pass · Clear and useful: pass · Tested: not-run · Featured standard: not-run
Per assistant
ChatGPT: not available (unavailable) · Claude: not available (unavailable) · Gemini: not available (unavailable) · Other AI: not available (unavailable)
no instruction text · safety review 2026-08-25