Verify Bug Skill for Claude

Verify Bug Skill

> [!IMPORTANT]

> Post-merge UAT verification workflow. Walks JIRA reproduce steps, performs comparative audits (Before/After), attaches evidence to JIRA, and transitions status on PASS.

Optional args: slug=<feature>, ticket=<id/url>, mode=interactive|autonomous|channel, channel=<id>, auto_continue=true|false, profile=business|hybrid|technical.

Instructions

When the user asks to perform this workflow, execute the following steps:

Verify-Bug — UAT Audit

Goal: Prove a bug fix works in the UAT environment via comparative Before/After evidence, then transition ticket status.

Input

/verify-bug <jira-url-or-key> [--baseline-image <url>]

Workflow

Step 0: Pre-flight & Data Gathering

> [!TIP]

> Sub-Agent Delegation: If your platform supports sub-agents (e.g., Claude, OpenCode, Gemini, Kiro), delegate steps 1-3 below to your JIRA Analyst sub-agent (e.g., @specialist-jira-analyst). If sub-agents are NOT supported (e.g., Antigravity, Windsurf), you must execute these steps yourself.

• Parse JIRA: Extract Market, Reproduce steps, and Expected Result.
• Resolve Markets: If multiple markets, prompt for scope (Full/Sample/Custom).
• Fetch Test Data: Call Confluence for Test data - <MARKET> UAT. Parse credentials and module-specific data (e.g., customer codes).
• Credential Check: Rule out expired accounts before starting sessions.
• Fallback: If Jira/Confluence MCPs are unavailable, request exported ticket/test-data text and continue with local evidence.

Step 1: Comparative Audit (Execution Phase)

For each market in scope:

• Environment Setup: Run the DNS probe from <SKILLS>/common/common-web-visual-testing/references/diagnostic-decoder.md; if it indicates VPN is required, connect VPN and retry.
• Named Session: Start playwright-cli -s={TICKET}-{MARKET} or Appium session.
• Walk Steps: Execute reproduction steps.
• Hover Discipline: Always hover the target element (warning, button, price) before screenshotting.
• Stability: Disable animations and mask dynamic fields (clocks, balances).
• Verdict Determination:
• PASS: End-state matches Expected Result.
• FAIL: End-state matches Actual Result or original bug screenshot.
• NEEDS-HUMAN: Deviates from both.

Step 2: Automated Failure Diagnostic

If the verdict is NOT PASS:

• Run Decoder: Load common-web-visual-testing; if synced references are available, consult <SKILLS>/common/common-web-visual-testing/references/diagnostic-decoder.md.
• Categorize: Is it a VPN NOT CONNECTED error? ACCOUNT BLOCKED? Or a genuine CODE REGRESSION?
• Label: Add the diagnostic label to the JIRA comment.

Step 3: Evidence & JIRA Sync

• Upload: Push screenshots as attachments to the JIRA ticket.
• Wiki Comment: Post a verdict comment using JIRA Wiki Markup (orientation-aware widths).
• Use 🟢 PASS / 🔴 FAIL badges.
• Embed the most diagnostic screenshot inline.
• Status Transition:
• If PASS: Ready for UAT → Ready for Production.
• If FAIL: → Reopened.
• Walkthrough:
• Use the Walkthrough Template below.
• Update project-local docs/srs/srs-walkthrough.md.

Runtime Contract

• Use for post-merge UAT verification of a bug fix against JIRA reproduce steps.
• Required inputs: JIRA URL/key or exported ticket text with reproduce steps and expected result.
• Return NEEDS-HUMAN only when the end-state deviates from both expected and original-bug behavior.

Handoff Payload

• slug, operator_profile (carried, not re-inferred), verdict (PASS/FAIL/NEEDS-HUMAN), walkthrough path, diagnostic label, outcome report, next workflow.

Blocking Questions

• Ask max 3 at a time with a recommended default and 2-3 options.

Artifact Templates

Walkthrough Template

# Walkthrough: [Name]

## Scope

## Acceptance Criteria

## Evidence

| Check   | Result              | Evidence   |
| ------- | ------------------- | ---------- |
| [check] | [PASS/FAIL/BLOCKED] | [evidence] |

## Risks

## Outcome Report
feature_status: implemented | blocked
requirement_trace: BRD-OBJ-* -> REQ-* -> AC-* -> SRS-* -> evidence
completed_evidence: []; missing_evidence: []; decision_needed: []; recommended_next_workflow: deploy-release | dev-fix

## Next Workflow
deploy-release | dev-fix

Cost Report

Call get_session_cost(workflow="verify-bug") before final handoff.

Anti-Patterns

• No Sequential Runs: Verify all markets in parallel.
• No Unnamed Sessions: Traceability depends on -s={TICKET}.
• No Mystery Failures: Always include the Diagnostic Decoder result in FAIL comments.
• No Orphan Comments: Clean up "temp media" comments after posting the final verdict.