qa-persona-init

qa-persona-init

Bootstrap persona-driven E2E testing on the current codebase. Drives Phases 1–4 of the qa-persona methodology with explicit stop-and-confirm gates.

For the full methodology, the rationale, and the scenario-category definitions, follow the parent skill at ${CLAUDE_PLUGIN_ROOT}/skills/qa-persona/SKILL.md. This skill is the launcher; the parent is the bible.

Inputs

$ARGUMENTS — optional flow name (used for scripts/qa/<flow>-e2e/ directory). If omitted, derive one from the codebase (e.g. marketplace, tasks, crm).

What it produces

docs/qa-persona/
  personas.md
  scenarios.md
  bugs.md
scripts/qa/<flow>-e2e/
  README.md
  lib.sh
  seed-personas.sh
  reset.sh
  run-all.sh
  scenarios/      (empty — populated as Phase 4 progresses)

Templates live at ${CLAUDE_PLUGIN_ROOT}/skills/qa-persona/templates/. Copy and adapt — never reference them at runtime.

Steps

Phase 1 — Codebase study

Read enough source to identify the value loop, actors, artifacts, and state transitions. Present back to the user in one paragraph.

STOP. Wait for the user to confirm or correct the value-loop summary before continuing. This is the contract for everything that follows.

Phase 2 — Persona proposal

Propose 3–5 named personas based on actors observed in the codebase. Use the schema from templates/persona-table.md. Fixed emails (persona-<role>@<co>.com) — never per-test bespoke users. Write to docs/qa-persona/personas.md.

STOP. Wait for user approval or edits before continuing.

Phase 3 — Scenario catalog

Author 5–8 scenarios across the seven categories (A–G). First batch MUST include at least one A, one D, AND one G. Status legend: ✅ / ⚠️ / 🔜. Write to docs/qa-persona/scenarios.md using templates/scenarios.md.

Phase 4 — Layer 1 scaffold

Create scripts/qa/<flow>-e2e/ from the bash templates (lib.sh, seed-personas.sh, reset.sh, run-all.sh). Replace FLOW_NAME and adapt API helpers to match the platform's actual auth + endpoints. Do NOT yet write per-scenario .sh files — those come from qa-persona-iterate or hand-authoring.

Red flags — do not skip

• Skipping Phase 1's stop-and-confirm gate.
• Picking personas "based on what feels right" instead of from observed actors.
• Deferring categories D or G past the first batch.
• Inventing a scenario catalog without the user agreeing on the value loop first.

After this

Run qa-persona-iterate <feature> whenever a feature ships, and qa-persona-walkthrough once Layer 1 is green.