smoke-test

name	smoke-test
description	Run smoke tests against a deployed environment to verify critical paths work

Skill: smoke-test

Load Context

1. Read $ARGUMENTS for the target environment or URL. If empty, ask the user.
2. Read the smoke test section of docs/RELEASE.md (if it exists)
3. Read all SPEC files under docs/modules/ to extract key API endpoints
4. Confirm the base URL for the target environment

Pre-checks

• If the target URL is unreachable (curl timeout) → abort: "Environment unreachable: . Check deployment status."
• If no environment identifier is provided and cannot be inferred → ask: "Please provide the target environment URL or name (staging/production)."

Steps

Step 1: Define smoke test scope

Determine test paths in priority order:

1. If a smoke-tests/ directory exists in the project root → use the defined test cases
2. If no dedicated tests exist → extract key endpoints from each module's SPEC API Contracts section and construct minimal requests

Critical path selection principles (pick 5–10):

• Service health check (GET /health or /api/health)
• Simplest happy path for core business flows (e.g. register, login, fetch resource)
• Any endpoint known to have caused issues at launch (inferred from notes or history)

Step 2: Execute checks

For each test path, run and record in this format:

[1/N] GET /health
      Expected: 200 OK
      Actual:   200 OK ✓

[2/N] POST /api/auth/register
      Expected: 201 Created
      Actual:   500 Internal Server Error ✗
      Response: {"error": "Database connection failed"}

Timeout per request: 10 seconds.

Step 3: Evaluate results

• All pass → proceed to Step 4 (success report)
• Failures exist:
  • Classify as "critical failure" (core functionality broken) or "non-critical failure" (edge case)
  • Critical failure → go to Error Handling immediately (trigger rollback)
  • Non-critical failure only → output a warning, ask the user whether to accept or trigger rollback

Step 4: Output report

Success:

✅ Smoke tests passed: <environment>

Checks: N passed / 0 failed
Duration: Xs

All critical paths are healthy.

Failure:

❌ Smoke tests failed: <environment>

Checks: M passed / K failed

Failed:
  ✗ [2/N] POST /api/auth/register
    Expected: 201, Actual: 500
    Response: {"error": "Database connection failed"}

Classification: critical failure
Recommended action: rollback immediately

Error Handling

Situation	Action
Critical path fails (production)	Output report, automatically trigger rollback, do not wait for user confirmation
Critical path fails (staging)	Output report, wait for user decision
All requests time out	Treat as environment unreachable — handle as critical failure
Auth endpoint returns 401 (test account issue)	Mark as "test configuration issue" rather than service failure — do not trigger rollback, ask user to check test credentials

Done When

• All critical paths have been tested (none skipped)
• Every check has an explicit pass/fail record
• On failure, the appropriate follow-up action has been triggered (rollback or user decision)
• Report has been shown to the user