test-executionlisted
Install: claude install-skill YevheniiaQA/qa-agent-toolkit
# Test Execution
The gap this fills: scenarios and cases exist, automation exists, and a manual
tester still has to assemble the environment, the accounts, the build number and
the right subset by hand before they can start. That assembly is the same work
every time, and getting it wrong is why a round of testing produces results
nobody can interpret.
## The execution pack
Write `04-results/<TICKET>/EXECUTE.md`. One page, everything needed to start,
nothing else. This file is the manual tester's front door: the root README and the
ticket README both point here.
```markdown
# Execute — PLAT-118 · Connect a Confluence workspace
**Build** 2026.09.14-a4f21 · **Environment** staging · **Estimated** 55 minutes
## Before you start
- URL: https://admin.platform.test.acme.com
- Accounts: `qa.admin@example.com` (superuser) · `qa.viewer@example.com` (viewer)
Passwords are in 1Password → QA → Platform staging
- VPN required: yes
- Do not run between 09:00 and 09:30 — staging deploys then
## Known defects — these cases are expected to fail
| Case | Defect | What you will see |
|---|---|---|
| TC-13 | DEF-PLAT-01 | Returns 200 and stores the record; the ticket requires rejection |
| TC-19 | DEF-PLAT-04 | Overwrites silently; the ticket requires 409 |
Mark them failed as normal — the record is what proves the defect is still open.
## Order
1. **Smoke** TC-01, TC-06 — if these fail, stop and report. Nothing else is meaningful.
2. **Core** TC-02 … TC-12 — the acceptance criteria
3.