manual-regression-web

name	manual-regression-web
description	Perform manual web regression using Playwright MCP tools (interactive). Required for web use-case changes. Keep evidence low-context (labels + paths only).

Manual Regression (Web via Playwright MCP)

When to use

Use when:

• Orchestrator flags use-case change affecting web UI, OR
• QA Gatekeeper requires manual regression, OR
• Automated E2E cannot be run in CI yet but validation is still required.

Manual regression is required for web use-case changes.

Inputs

• UC docs: /docs/use-cases/UC-###-*.md
• Running environment (local dev server and/or docker-compose)
• Playwright MCP tool access

Outputs

• Manual validation of UC acceptance criteria using Playwright MCP tools
• MCP proof label recorded: playwright-mcp
• Any added/updated Playwright specs recorded under tests_added[] (if created)

Rules

• Do not paste tool logs or long transcripts.
• Validate observable outcomes aligned to UC acceptance criteria.
• If the environment cannot be run (missing secrets/ports/credentials), create blockers via MCP and keep completion gated.

Step 1 — Ensure system is running

Preferred:

• make up (if docker-compose stack)
• or make dev / make run (repo convention) Confirm:
• Web app reachable
• Backend reachable if required for the flow

If not reachable:

• fix runtime/compose issues first (blocking)

Step 2 — Create a UC checklist (compact)

For each affected UC:

• list 3–8 checkpoints max:
  • navigation start
  • key input(s)
  • key output(s)
  • one critical error/edge path (if relevant)

Step 3 — Execute manual regression with Playwright MCP

Using Playwright MCP tools:

• Navigate the flow
• Capture failures precisely (what step, what expected, what happened)
• If failure found:
  • create or update automated Playwright spec if feasible
  • otherwise create a task/blocker describing the failure and required fix

Keep interaction tight:

• Prefer direct steps; avoid exploratory wandering
• Repeat only the minimal flow necessary to confirm fix

Step 4 — Record outcomes (low-context)

Do NOT store logs. Do:

• note which UC IDs were manually validated (in task notes if needed)
• record proof label playwright-mcp in MCP manual_regression[]

If the manual regression depended on specific local setup:

• add a short note in /docs/testing/how-to-run.md (only if it’s stable and reusable)

Step 5 — Completion proof (MCP)

When marking task done:

• manual_regression[] must include: playwright-mcp
• tests_added[] includes any new/updated spec file paths
• changed_files[] includes any config/doc paths changed

Suggested required_gates[]

• For web use-case change: manual-regression (and typically e2e) plus baseline lint/build/ci