Verify Task

Start an existing Task as a QA engineer.

Use the Gherkin scenarios as the executable test script. Each scenario is a concrete test case with Given/When/Then steps to run against the implementation.

Read references/qa-verdict-guide.md before you start. It defines the status symbols, verdict options, and the expected Test Mapping table format used throughout this skill.

Process

0. GitHub CLI setup

Run steps 1–2 of ../references/gh-setup.md (install check and auth check). If gh is not installed or not authenticated, stop. Extensions are not required for this skill.

If this skill was invoked from wtf.implement-task or another skill that already ran gh-setup this session, skip this step.

1. Identify the verification scope

Call AskUserQuestion (per ../references/questioning-style.md):

• question: "Are you verifying a single Task or a full Feature?"
• header: "Scope"
• options:

- Single Task → verify one Task's Gherkin scenarios - Full Feature → verify all Tasks under a Feature using its sub-issues

If Single Task:

Call AskUserQuestion (per ../references/questioning-style.md):

• question: "Which Task are you testing?"
• header: "Task"
• options: from recent open issues labeled task or implemented

Walk Task → Feature per ../references/spec-hierarchy.md to extract Gherkin, Contracts, Edge Cases, Test Mapping, DoD (Task) and ACs / edge cases for additional probe scenarios (Feature).

Apply the absent-label gate from ../references/lifecycle-labels.md for the implemented label on the Task.

Recommended skill: wtf.implement-task. Header: Implement first?.

On Implement first → follow wtf.implement-task and pass the Task number as context. On Verify anyway → continue.

If Full Feature:

Call AskUserQuestion (per ../references/questioning-style.md):

• question: "Which Feature are you verifying?"
• header: "Feature"
• options: from open feature issues

Fetch all sub-issues of the Feature using gh sub-issue list <feature_number> per the cookbook in ../references/gh-setup.md. This returns the authoritative list of Tasks. Do not search by label or title matching.

Sub-phase the task list. Apply the file-conflict coloring algorithm in ../references/conflict-graph.md to partition the tasks into conflict-free sub-phases. Tasks within a sub-phase can run in parallel worktrees. Sub-phases run in sequence.

Spawn one sub-agent per task, per sub-phase. Apply the rules in ../references/subagent-protocol.md for every Agent call. Specifically for Full Feature mode:

• Read skills/wtf.verify-task/SKILL.md at spawn time and paste steps 3–9 into each sub-agent prompt (inline instructions — sub-agents cannot load this skill by name).
• Override interactive prompts as follows:

- "Test surface" confirmation (step 3) → skip, proceed with all found scenarios - Per-scenario pass/fail (step 4) → return as NEEDS_INPUT to the orchestrator - Bug-filing prompts → defer to the aggregated step 9

• Mandatory label (non-skippable): gh issue edit <task_number> --add-label "verified" after verification completes.

Wait for all sub-agents in a sub-phase to complete. Resolve any NEEDS_INPUT responses before you start the next sub-phase.

After all sub-phases finish, aggregate into a feature-level summary. Include total tasks and pass/fail/blocked counts. Present it.

2. Load the QA steering document

Load docs/steering/QA.md per the strict consumer-side load in ../references/steering-doc-process.md (recommended skill: wtf.steer-qa). Apply its test strategy, coverage thresholds, definition of done, and known flaky areas silently throughout this session.

3. Establish the test surface

From the Task, extract and present:

• All Gherkin scenarios (these are the test cases)
• The contracts (request/response schemas to verify against)
• Edge Cases & Risks (additional scenarios to probe)
• Observability requirements (logs, metrics, alerts to verify)

Call AskUserQuestion (per ../references/questioning-style.md):

• question: "I found [n] Gherkin scenarios and [m] edge cases to cover. Does this match what you expect?" (replace [n] and [m] with actual counts)
• header: "Test surface"
• options:

- Yes — that's everything → proceed to testing - There are more scenarios → add them first

4. Walk through each Gherkin scenario

For each scenario, one at a time, call AskUserQuestion (per ../references/questioning-style.md):

1. Present it as a concrete test case. Restate the Given/When/Then in plain language.
2. Ask "Did this scenario pass?" — header Result:

- Yes ✅ → mark ✅ in the running Test Mapping table. Set bug filed to — - No ❌ → record failure (see below) - Blocked 🚫 → record blocker (see below) - N/A or Conditional ⚠️ → record condition (see below)

On No ❌: call AskUserQuestion (per ../references/questioning-style.md): - question: "What actually happened?" - header: "Failure details" - options: from plausible failure modes inferred from the scenario (e.g. "No error shown", "Wrong data returned") Record findings with repro steps. Then call AskUserQuestion (per ../references/questioning-style.md): - question: "Would you like to file a bug report now?" - header: "File bug?" - options: - File now → run wtf.report-bug immediately with the task number and scenario details (default). Mark bug filed as yes - Continue and file later → defer. Mark bug filed as no

On Blocked 🚫: call AskUserQuestion (per ../references/questioning-style.md): - question: "What dependency or environment issue prevented testing?" - header: "Blocker" - options: from common blockers inferred from the task context (e.g. "Missing test environment", "Depends on unmerged task") Set bug filed to —.

On N/A or Conditional ⚠️: call AskUserQuestion (per ../references/questioning-style.md): - question: "Is this N/A, or does it pass only under a condition?" - header: "Condition" - options: - N/A — not applicable → this scenario does not apply - Conditional — specify the condition → passes only under a specific circumstance

Record appropriately. Set bug filed to — (track the condition separately).

1. After you record the result, immediately update the Task issue with the current state of the Test Mapping table. Do not wait until all scenarios are done.

The table must include a Bug Filed column:

The running Test Mapping table format (update after every scenario):

``bash python3 .wtf/gh-body.py read <task_number> # prints a temp path ``

Programmatically replace the Test Mapping table section in the printed temp file using the Write or Edit tool. Preserve all other sections unchanged. Then push (see ../references/gh-body-helper.md):

``bash python3 .wtf/gh-body.py edit <task_number> --body-file "<path-from-read>" ``

1. Keep a running tally. After updating, confirm: "Updated. Moving to next scenario..."

5. Probe the edge cases

For each Edge Case listed in the Task (and the parent Feature), one at a time:

1. Derive a concrete test action from the edge case description.
2. Call AskUserQuestion (per ../references/questioning-style.md):

- question: "Did this edge case pass?" - header: "Result" - options: Yes ✅ / No ❌ / Blocked 🚫 / N/A

On No ❌: call AskUserQuestion (per ../references/questioning-style.md): - question: "What actually happened?" - header: "Failure details" - options: from plausible failure modes inferred from the edge case Record findings with repro steps. Then ask to file a bug report as in step 4.

1. After each result, update the Task issue. Append an Edge Cases section (or update it if present) with the same table format used in step 4.

6. Verify observability

For each item in the Observability section (logs, metrics, alerts), one at a time, call AskUserQuestion (per ../references/questioning-style.md):

1. Ask "Was this observability item present and correct?" — header Result:

- Yes ✅ → present and correct - No ❌ → missing or incorrect - N/A → not applicable to this task

1. Record the result. On ❌, ask for details and offer to file a bug report as in step 4.
2. After each result, update the Task issue with an Observability Results section.

7. Finalize results and post QA summary

Apply strict STE per ../references/ste-writing.md before you write any durable body (QA summary comment and free-text issue updates).

The Test Mapping table has been updated after each scenario (step 4). Now do a final update. Check off DoD items that passed. Leave failing ones unchecked.

python3 .wtf/gh-body.py read <task_number>       # re-fetch; prints a fresh temp path

Programmatically update the DoD checklist in the printed temp file using the Write or Edit tool. Then push:

python3 .wtf/gh-body.py edit <task_number> --body-file "<path-from-read>"

Post a QA summary comment:

# Write the QA summary to a temp file with the Write tool; $COMMENT is that path.
python3 .wtf/gh-body.py comment <task_number> --body-file "$COMMENT"

The QA summary must include:

• Total scenarios tested and pass/fail/conditional count
• Any findings with repro steps
• Conditional passes: list each ⚠️ scenario with its required condition
• Clear verdict: ✅ Ready for merge / ❌ Needs fixes / ⚠️ Conditional pass (list conditions)

If the verdict is ✅ or ⚠️, add the verified lifecycle label:

gh issue edit <task_number> --add-label "verified"

Print the updated Task issue URL.

8. Offer to open a PR and close the issue

If the verdict is ✅ or ⚠️, call AskUserQuestion (per ../references/questioning-style.md):

• question: "Task verified. What would you like to do next?"
• header: "Next step"
• options:

- Open PR now → follow the wtf.create-pr process, passing the Task number in as context (recommended) - Skip for now → exit. Open the PR later yourself

Closing policy: Issues are only "closed as completed" via a merged PR that contains Closes #<n>. Never call gh issue close <n> for completed work. Direct closes are reserved for:
- gh issue close <n> --reason "not planned" — will not implement
- gh issue close <n> --reason "duplicate" — duplicate of another issue

9. Offer bug reports for remaining failures

Check all result tables (Gherkin scenarios from step 4, edge cases from step 5, observability from step 6). Find all rows where Result is ❌ and Bug Filed is no. These are the unfiled failures.

If none exist, skip this step entirely.

If unfiled failures exist, present them as a numbered list. Then call AskUserQuestion (per ../references/questioning-style.md):

• question: "[n] failing scenario(s) without a bug report. How would you like to handle them?" (replace [n] with the actual count)
• header: "File bugs?"
• options:

- File separately → spawn one sub-agent per failing scenario in parallel using the Agent tool, each running the report-bug fast path. Apply ../references/subagent-protocol.md. Wait for all sub-agents to complete before exiting (default). - File combined → follow the wtf.report-bug process once, passing in the task number and all failing scenarios together. - Skip → exit without filing reports.