smoke-test

name	smoke-test
description	Diagnose an integration/E2E test log, classify failures by severity, and batch-create issues. Use when the user says "smoke test", "diagnose log", "triage test log", or provides a test log.
argument-hint	<path to log file or paste log>
user-invocable	true
allowed-tools	Bash, Read, Grep, Glob, TodoWrite
model	sonnet

/smoke-test — Integration Test Log Diagnosis & Triage

Analyze an integration or E2E test log, diagnose each failure against the source code, and batch-create issues for new findings.

Steps

1. Load the log from $ARGUMENTS (file path or inline text). If no argument, ask for the log file path.

2. Parse all failures and warnings — extract:

   • Error messages and their context (5 lines before/after)
   • Components or modules involved
   • Function calls or API endpoints that failed
   • Network, timeout, or connection errors
   • Any stack traces or crash indicators

3. Diagnose each failure — for every distinct error:

   • Search the source code for the function/path that produced the error
   • Identify the root cause (wrong parameter, API response parsing, timing, config issue)
   • Classify severity:
     • Critical: Core functionality failure, data corruption, security issue
     • Medium: Degraded functionality, fallback triggered, non-fatal error
     • Low: Cosmetic, logging, or non-functional issue

4. Check for existing tracked issues — for each finding:

   • Check docs/LINEAR_SNAPSHOT.md or docs/KNOWN_ISSUES.md for existing entries
   • Search by keyword: component name, error message, function name
   • Mark as "EXISTING: [ID]" if found, "NEW" if not

5. Create issues for NEW findings — use the project's issue tracking:

   • If using Linear: create via linear issue create CLI
   • If using local tickets: create via /create-ticket
   • Set priority: Critical=High/Urgent, Medium=Medium, Low=Low
   • Include in description: error message, affected file:line, suspected root cause, log context

6. Output summary table:

   ## Smoke Test Triage: [date/log name]
   
   | # | Issue | Severity | Root Cause | File:Line | Tracking |
   |---|-------|----------|------------|-----------|----------|
   | 1 | ... | Critical | ... | src/...:NN | NEW: [ID] |
   | 2 | ... | Medium | ... | src/...:NN | EXISTING: [ID] |
   
   ### Statistics
   - Total failures: N
   - Critical: N | Medium: N | Low: N
   - New issues created: N
   - Already tracked: N
   

Critical Rules

• Diagnose against source, not assumptions. Every root cause must reference actual code with file:line.
• Do not fix anything. This skill is diagnosis and triage only.
• Do not create duplicate issues. Always check existing tracking first.
• Group related errors. Multiple log lines from the same root cause = one issue, not many.
• Include log context in issue descriptions. Future fixers need the evidence.

Notes

• Use TodoWrite to track multiple failures being investigated in parallel
• If the log is very large, focus on Critical and Medium issues first
• After this skill completes, use /fix-issue [ID] to fix individual issues