test-estimationlisted
Install: claude install-skill YevheniiaQA/qa-agent-toolkit
# Test Estimation
Sprint planning is where QA either gets time or gets blamed later for not having
had any. The skill here is not producing a number — it is producing a number
somebody can argue with, because an estimate whose reasoning is invisible gets
halved in the room.
## Estimate from the decomposition, not the title
A story title carries no information about test effort. Run
`requirements-decomposer` first, or estimate from an existing decomposition. The
requirement count, the number of input classes, and the number of open questions
predict effort far better than story points do.
```markdown
## PLAT-118 — QA estimate
**Scenarios expected**: 22–28
12 functional requirements × ~1.6 scenarios (positive + negative + boundary)
3 non-functional requirements requiring their own setup
**Effort**
| Activity | Hours | Basis |
|---|---|---|
| Requirements analysis and questions | 1.5 | 12 requirements, 4 already ambiguous |
| Scenario design | 3 | 25 scenarios, 2 needing decision tables |
| API automation | 5 | 18 API scenarios, existing fixtures reusable |
| UI automation | 7 | 7 UI scenarios, 4 new page objects, 3 test IDs to request |
| Debugging and stabilisation | 4 | ~30% of authoring time, from the last three stories |
| **Total** | **20.5** | |
**Confidence**: medium. Rises to high once Q-02 and Q-05 are answered.
```
The basis column is what survives challenge. "Five hours" invites a haggle;
"eighteen scenarios against existing fixtures" invites a conversati