qa-persona-iterate

qa-persona-iterate

Run after every feature ship. Closes the loop the methodology promises: catalog grows monotonically, walkthrough is appended (never rewritten), bug tracker stays current.

Inputs

$ARGUMENTS — free-form description of the feature that just shipped. If omitted, infer from the current branch diff (git diff main...HEAD --stat) and recent commit messages.

Pre-checks

1. docs/qa-persona/scenarios.md exists. If not → tell the user to run qa-persona-init first and stop.
2. Detect the flow name from scripts/qa/<flow>-e2e/.
3. Detect the current Run # by counting ^## Run # headers in the latest walkthrough-RUN-*.md (default: next is current+1).

Steps

1. Identify affected scenarios

Read docs/qa-persona/scenarios.md. For each existing scenario ID, decide whether the feature touches it. Output a table:

Scenario	Affected?	Why
A1	yes	feature changes the happy-path lifecycle
D1	no	auth model unchanged

2. Propose new scenarios

For each of the seven categories, ask: does this feature need a NEW scenario in this category? Lean toward yes for A (happy path), D (auth/isolation), and G (cross-cutting / audit) — those decay fastest.

Propose new scenario IDs continuing the existing numbering (e.g. if last A is A2, propose A3-<feature-slug>.sh). Output a second table:

New ID	Category	Title	Why
A3	Happy path	multi-assignee task lifecycle	feature: multi-assignee tasks
E2	Multi-actor	concurrent assignee claim race	feature: multi-assignee tasks

STOP. Show both tables to the user. Wait for approval before scaffolding.

3. Scaffold runners

For each approved new scenario:

• Copy ${CLAUDE_PLUGIN_ROOT}/skills/qa-persona/templates/scenario.sh to scripts/qa/<flow>-e2e/scenarios/<ID>-<slug>.sh.
• Replace placeholders: scenario ID, title, persona emails, action sequence skeleton.
• Mark the scenario in scenarios.md with status 🔜 → ⚠️ once authored, ✅ once it passes.

4. Stage Run #N skeleton

Find the existing walkthrough-RUN-1.md (the journal — filename never changes). Append (do not rewrite):

## Run #N — <feature> (<YYYY-MM-DD>)

**What changed since Run #N-1:**
- <bullet from $ARGUMENTS>

**Personas re-used:** <list>
**Scenarios walked:** <new + affected IDs>

### Findings

✅ **Fixed since Run #N-1:** <bug IDs from bugs.md, or "—">

🐛 **New bugs surfaced:** [filled in during qa-persona-walkthrough]

### BR coverage update

[table]

### Verdict update

[one-paragraph delta]

5. Update bug tracker delta

In docs/qa-persona/bugs.md, add a new ### Run #N (<date>) section under "Per-run delta" with placeholders New bugs: — and Fixed: —. The walkthrough fills these in.

Output

• New scripts/qa/<flow>-e2e/scenarios/*.sh files.
• Updated docs/qa-persona/scenarios.md with new rows.
• Appended ## Run #N skeleton in the walkthrough doc.
• New ### Run #N block in bugs.md.

Red flags — do not do

• Rewriting Run #N-1 instead of appending Run #N.
• Skipping category G when the feature touches a state mutation (audit-log scenarios are the easiest to forget and the ones legal cares about).
• Auto-running the new scenarios without confirming with the user — author first, run via qa-persona-walkthrough or run-all.sh second.