Builds stakeholder-readable scripted manual test cases from a feature spec in four formats: a step-table (preconditions / steps / expected result / actual / pass-fail / notes) for spreadsheet review, a Gherkin Given/When/Then format for BDD-aware teams, a business-language UAT script with acceptance-criteria mapping and contractual sign-off (references/uat-format.md), and a one-line-per-item execution checklist for smoke / on-call / bug-bash / compliance sweeps (references/checklist-format.md). Each script is self-contained (no implicit team knowledge), single-scenario (one happy + N edge per script), and includes the data setup the tester needs without being a developer. Use when a feature can't be (or shouldn't be) fully automated and a human tester needs an executable script or checklist - UAT sign-off rounds, regression baselines, certification testing, deploy smoke checklists, exploratory follow-up scripts.
94
88%
Does it follow best practices?
Impact
95%
0.98xAverage score across 10 eval scenarios
High
Do not use without reviewing
We deploy the payments dashboard on Tuesdays and Thursdays. The engineer on call
is supposed to walk the wiki page wiki/deploy-smoke.md afterwards. In the last
eleven deploys it was walked twice. Both times it took the engineer close to
fifty minutes, and on one of them they were interrupted at item 19 and never went
back.
The page has accumulated for two years. It is one long undivided list, several entries are small projects in their own right, and a good number tell the engineer to look at something without saying what a good result looks like. One entry cannot be completed at deploy time at all.
There is a second problem that only showed up after the January incident. The
page is edited in place by whoever ran it last, so when we tried to establish
what had actually been verified on the afternoon the outage began, there was no
record - only the current version of the page. Two earlier attempts at a shorter
list are sitting in docs/checklists/ as prod-smoke-v1.md and
prod-smoke-v2.md; leave both alone.
What we want is a sweep the on-call can genuinely finish before the deploy window closes, and a completed run that is still readable six months later.
Produce one markdown document at exactly docs/checklists/prod-smoke-v3.md
containing:
Out of scope: automating any of it, editing the wiki page, and touching
prod-smoke-v1.md or prod-smoke-v2.md.
Extract the following files before beginning.
=============== FILE: wiki/deploy-smoke.md ===============
Edited by whoever ran it last. Keep it current.
=============== FILE: docs/prod-accounts-and-data.md ===============
The on-call uses a ring-fenced sandbox merchant that sits in production but settles to a dummy ledger. Nothing done under it moves real money, and nothing outside it should be touched during a deploy sweep.
| Item | Value |
|---|---|
| Login | oncall.smoke@paylane.example |
| Password | 1Password entry prod-smoke-login |
| Merchant | Sandbox Merchant (SBX-9) in the merchant switcher |
| Second merchant | Sandbox Merchant (SBX-10) - for switcher checks |
| Known transaction | reference TXN-SBX-000117, 42.00 USD, settled |
| Known payout | PO-SBX-0042, 1,204.75 USD, status Paid |
| Report to spot-check | "Daily settlement" - has data every day |
| Notification inbox | oncall.smoke@paylane.example (real mailbox) |
The release version is printed in the footer of every page as build: <sha>
and is also on the deploy message in #deploys.
The nightly settlement batch runs at 02:10 UTC. Nothing that depends on it can be verified during a daytime deploy.
Refunds on the sandbox merchant are safe. Refunds on any other merchant are real customer money.