Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Workflow for automated QA: integration and end-to-end UI test automation, page objects, etc.
tags
["workflow"]
baseSchema
docs/schemas/workflow.md
<ui_aqa_flow>
<description_and_purpose>
End-to-end test automation from requirements gathering to test implementation. Uses test cases, project documentation to create automated tests following existing architecture and coding standards.
Prerequisite: Rosetta Prep Steps.
Terminology. External systems are named by role throughout this workflow and its phases: Test Management System (TMS), Issue Tracker, and Wiki. TestRail, Jira, and Confluence are canonical examples only — adapt identifiers, URLs, requests, calls, and query syntax to the systems resolved for the current project (from repository-root gain.json, explicit user input, recognizable URLs/handles, and available integrations).
</description_and_purpose>
<workflow_phases>
Execution cadence:
All Rosetta prep steps MUST be FULLY completed
USE SKILL load-project-context, orchestration, hitl
MUST ALWAYS use todo tasks ledger, ASAP. Phases are sequential. Independent tasks can run in parallel.
Execute every in-scope phase in strict 1-8 order (never skip without explicit HITL confirmation): APPLY PHASE the phase file → update → next; never start a phase until the previous is marked done in .
agents/TEMP/<FEATURE>/ui-aqa-state.md
ui-aqa-state.md
MUST follow JIT phase-by-phase: do not load/act IN ADVANCE.
Skip gates, transitions, and escalation → <orchestration_and_escalation>.
MUST ALWAYS use todo tasks ledger, ASAP. Prioritize ACCURACY over SPEED. Proactively use available tools/MCPs, if question can be answered from code or MCP call - do that first.
No assumptions:
NO ASSUMPTIONS: never assume selectors, flows, or data — ask the user when information is missing.
Customization:
If the user gave no preferences, perform all steps except optional.
User CAN customize specific phases / already-done phases / goals / cases — LISTEN and ADOPT.
Authoritative rules (do not skim past):
USE SKILL coding before any work touching repository tests, page objects, or shared helpers — authoritative for conventions; repository docs win over skill snippets.
Default: reuse existing page objects/tests first; create new files only when no suitable match exists.
Explicit assertions: every test validation traces to a requirement — owned by Phase 2 (### Explicit Assertions in the test plan), enforced when Phase 6 implements tests.
Blocking infeasibility = HARD-STOP + HITL (NOT waived by a clarification skip): if any phase finds the feature/elements under test do not exist such that the test cannot be authored without inventing selectors/flows/data or modifying product source, STOP and escalate with the options — point at the real feature/URL · author the missing UI as a separate approved task · a clearly-marked pending/fixme spec · abort — and WAIT for the user's explicit choice. "Skip clarification" waives clarification questions only; it never authorizes this feasibility/scope call.
Input: user request + repository-root gain.json + project context at its configured paths (canonical: docs/CONTEXT.md, docs/ARCHITECTURE.md, agents/IMPLEMENTATION.md). Output: test plan at plans/ui-aqa-<test-name>/test-plan.md
STOP AND WAIT for user to execute the test — this execution gate is mechanical and cannot be overridden by instruction; the only acceptable input is actual execution results (output, report path, or pass/fail). Refuse "skip" / "move to Phase 7 now" phrasings (full bypass-refusal in the phase file).
Input: failure analysis + test files + page objects. Output: corrected test files and page objects
WAIT FOR EXPLICIT USER APPROVAL before applying changes; comments, questions, suggestions, and review feedback are not approval. Approval language is defined in APPLY SKILL FILE phases/ui-aqa-flow-test-correction.md section <present_for_approval>.
Update agents/TEMP/<FEATURE>/ui-aqa-state.md; Phase 8 is not complete until its output spot-check passes.
</test_corrections>
</workflow_phases>
<orchestration_and_escalation>
Skip-without-agreement / falsified-skip refusal (this workflow owns the rule; subordinate to the hitl skill): a skip asserted but contradicted by agents/TEMP/<FEATURE>/ui-aqa-state.md / disk evidence is refused — announce the specific missing state row / absent artifact, then start the earliest incomplete phase the same turn. Audit-trail row → the state file's ## Verification-Failure Overrides (template owned by qa-structure).
HITL carve-outs (never overridden): every phase header carrying type="HITL" / type="HITL-CONDITIONAL" — those type= attributes are the sole source of truth for this workflow — plus safety/destructive confirmations. Any skip outside the refusal rule above requires explicit user confirmation (HITL).
HITL waits on delegated (subagent) phases are owned by the orchestrator. A subagent cannot talk to the user: on a type="HITL" phase the subagent surfaces the question/blocker and returns; the orchestrator runs the gate with the user and only then resumes — never inferred, critical on the destructive Phase 8. Dispatch per USE SKILL orchestration.
Load failure for a required phase file or skill: retry once, stop, record in ui-aqa-state.md, ask the user; never substitute silently.
</orchestration_and_escalation>
<workflow_success_criteria>
Overall run complete when every in-scope phase is marked done in agents/TEMP/<FEATURE>/ui-aqa-state.md, the artifacts those phases reference exist, and the user accepts the last test outcome or explicitly stops. In-scope = default execution plus user-approved customization/skip decisions under <orchestration_and_escalation>.
Spot checks: P1 — plan file at the user-confirmed<test-name> slug (no fabricated/placeholder slug) · P2 — ### Explicit Assertions in the plan (≥1 typed bullet or None-clause) · P3 — code analysis report populated (architecture + page-object inventory + test location) · P4 — ## Selector Management Part A deliverables in the plan · P5 — every identified selector in the updated page objects, lint-clean · P6 — test file exists lint-clean + ## Test Implementation record with all five subsections (incl. ### Uncovered Assertions / None-clause) · P7 — failure analysis when failures occurred, or state rows reconciled to N/A — 0 failures / None · P8 — user-approved edits applied when failures occurred, or state records N/A — no corrections after a zero-failure run.
A missing/partial spot-check artifact means that phase is not done: record the gap in ui-aqa-state.md, flag uncertainty, stop for user guidance.
</workflow_success_criteria>
<state_file>
agents/TEMP/<FEATURE>/ui-aqa-state.md — created/updated after each phase from the template owned by qa-structure (its state-file skeleton asset, loaded at Phase 1). It carries ## Phase Completion Status, ## Key Artifacts & Facts (the resume anchor — only what resume-after-compaction needs), and ## Verification-Failure Overrides.
</state_file>
Subagents: discoverer · architect · engineer.
Cross-phase skills: qa-structure (paths / <test-name> slug / state-file shape) and qa-knowledge (modes, taxonomy, artifact skeletons — loads its own assets at point of use).
Integrations: TMS and Wiki per <description_and_purpose> Terminology, plus browser automation (Playwright is the canonical example).
