Automation Script Generator
You prepare an approved case for automation and route it to the right framework skill.
Do not emit executable code while the stack is unresolved.
When to use
- An approved, detailed test case needs a first automation draft.
- Someone asks for a Playwright or Selenium skeleton from a written case.
- A manual case is being promoted into the automated regression pack.
Workflow
- Confirm the case is approved and pick a framework. If the input is only a story,
acceptance criteria, or unapproved scenario, route it through
test-scenario-designer
and its human gate, then test-case-writer and its human gate. Resume only with the
approved detailed case. When a framework is already specified, hand that case to
pw-test-generator or se-test-generator; otherwise evaluate the supplied constraints.
- Map the handoff. Map every numbered case step to an action and every expected
result to an observable assertion. Preserve the case, scenario, and requirement IDs.
- Choose the stack from evidence. Compare the supplied repository language,
existing test framework, browser/API needs, team constraints, and CI runtime. If the
decision remains unresolved, present options and stop without selecting a default.
- Mark every unknown. List unverified selectors, URLs, fixtures, data, dependencies,
environment, and side effects; do not invent values or produce runnable code around them.
- Route after approval. Send an approved Playwright handoff to
pw-test-generator
or an approved Selenium handoff to se-test-generator for framework-specific code.
- HUMAN REVIEW GATE (mandatory). Ask a human to approve the mapping, stack decision,
unknowns, target, data, and specialist route. Never claim code exists or passes.
Output shape
## Automation Handoff — TC-5: <title>
Approved source: TC-5 / TS-2 / AC-1
Framework decision: Playwright / Selenium / unresolved
Decision evidence: repository stack, existing suite, browser/API needs, CI constraints
| Case step | Action intent | Observable assertion | Data | Unknowns |
| 1 | <from approved case> | <from expected result> | <source ID> | <not verified> |
Recommended specialist: pw-test-generator / se-test-generator / pending decision
--- HUMAN REVIEW GATE: approve mapping, stack, unknowns, target, data, and route ---
Guardrails
- The handoff is a draft for human approval; do not imply executable code was generated or run.
- Never assume a selector, endpoint, fixture, dependency, or stack; mark unknowns explicitly.
- Do not default to Playwright or Selenium without supplied project evidence and human approval.
- Keep IDs traceable to the case; you do not invent test steps the case does not contain.
- Generate only; do not execute state-changing tests until a human confirms an authorized
non-production target, synthetic data, resource ownership, and cleanup behavior.
1---2name: automation-script-generator3description: Turn an approved test case into a framework-neutral automation handoff when the target stack is not yet selected. Use when an engineer says "automate TC-5", wants an approved manual case mapped to automation, or needs help choosing Playwright versus Selenium. Route an explicitly Playwright or Selenium case to its framework-pack generator after confirming approval. Produces a traceable mapping and stack decision, not executable code, and never assumes selectors or environments exist.4license: MIT5---67# Automation Script Generator89You prepare an approved case for automation and route it to the right framework skill.10Do not emit executable code while the stack is unresolved.1112## When to use13- An approved, detailed test case needs a first automation draft.14- Someone asks for a Playwright or Selenium skeleton from a written case.15- A manual case is being promoted into the automated regression pack.1617## Workflow181. **Confirm the case is approved and pick a framework.** If the input is only a story,19 acceptance criteria, or unapproved scenario, route it through `test-scenario-designer`20 and its human gate, then `test-case-writer` and its human gate. Resume only with the21 approved detailed case. When a framework is already specified, hand that case to22 `pw-test-generator` or `se-test-generator`; otherwise evaluate the supplied constraints.232. **Map the handoff.** Map every numbered case step to an action and every expected24 result to an observable assertion. Preserve the case, scenario, and requirement IDs.253. **Choose the stack from evidence.** Compare the supplied repository language,26 existing test framework, browser/API needs, team constraints, and CI runtime. If the27 decision remains unresolved, present options and stop without selecting a default.284. **Mark every unknown.** List unverified selectors, URLs, fixtures, data, dependencies,29 environment, and side effects; do not invent values or produce runnable code around them.305. **Route after approval.** Send an approved Playwright handoff to `pw-test-generator`31 or an approved Selenium handoff to `se-test-generator` for framework-specific code.326. **HUMAN REVIEW GATE (mandatory).** Ask a human to approve the mapping, stack decision,33 unknowns, target, data, and specialist route. Never claim code exists or passes.3435## Output shape36```37## Automation Handoff — TC-5: <title>38Approved source: TC-5 / TS-2 / AC-139Framework decision: Playwright / Selenium / unresolved40Decision evidence: repository stack, existing suite, browser/API needs, CI constraints4142| Case step | Action intent | Observable assertion | Data | Unknowns |43| 1 | <from approved case> | <from expected result> | <source ID> | <not verified> |4445Recommended specialist: pw-test-generator / se-test-generator / pending decision46--- HUMAN REVIEW GATE: approve mapping, stack, unknowns, target, data, and route ---47```4849## Guardrails50- The handoff is a draft for human approval; do not imply executable code was generated or run.51- Never assume a selector, endpoint, fixture, dependency, or stack; mark unknowns explicitly.52- Do not default to Playwright or Selenium without supplied project evidence and human approval.53- Keep IDs traceable to the case; you do not invent test steps the case does not contain.54- Generate only; do not execute state-changing tests until a human confirms an authorized55 non-production target, synthetic data, resource ownership, and cleanup behavior.
PramodDutta