kinginyellows Test Conventions
What It Does
Reference conventions for the yellow-browser-test plugin's testing workflow. Defines the config schema, report format, severity levels, and safety rules.
When to Use
Use when yellow-browser-test commands or agents need the config schema, severity classification, or credential handling rules.
Usage
This skill is not user-invocable. It provides shared context for the yellow-browser-test plugin.
Config Schema
Config is stored at .claude/yellow-browser-test.local.md as YAML frontmatter +
markdown notes.
---
schema: 1
generated: '2026-02-13T10:30:00Z'
devServer:
command: 'npm run dev'
port: 3000
baseURL: 'http://localhost:3000'
readyTimeout: 60
readyPath: '/'
auth:
type: 'email-password'
loginPath: '/login'
credentials:
email: '$BROWSER_TEST_EMAIL'
password: '$BROWSER_TEST_PASSWORD'
formFields:
email: 'email'
password: 'password'
routes:
- path: '/dashboard'
name: 'Dashboard'
auth: true
---
Required fields: schema, devServer.command, devServer.baseURL
Severity Classification
| Severity | Criteria | Example |
|---|---|---|
| critical | App crashes, data loss, security hole | Unhandled exception on form submit |
| major | Feature broken, console error, blank section | Settings page renders empty |
| minor | Slow load, UI glitch, non-blocking error | Button misaligned, 5s load time |
| cosmetic | Visual imperfection, typo | Font inconsistency, extra whitespace |
Credential Handling Rules
- Never store actual credentials — config references env var NAMES only
(
$BROWSER_TEST_EMAIL) - Auth state file —
.claude/browser-test-auth.jsonmust be gitignored - Missing vars — If env vars are not set, error with: "Set {VAR_NAME} in your environment or .env.local"
- PID file —
.claude/browser-test-server.pidmust be gitignored - Env var naming — Use
BROWSER_TEST_*prefix for all test credentials (e.g.,BROWSER_TEST_EMAIL,BROWSER_TEST_PASSWORD) - Never log credentials — Mask credential values in reports with
*** - Auth state expiry — If tests fail with auth errors, delete
.claude/browser-test-auth.jsonand re-run setup to refresh auth state
Safe Exploration Rules
During autonomous exploration, the test-runner agent MUST:
- Follow safety rules from agent-browser-patterns skill
- Limit depth — Max 3 levels from starting page
- Limit time — Max 10 minutes per exploration session
- Save incrementally — Write partial results every 5 routes to prevent data loss
Report Template
Reports are written to test-reports/YYYY-MM-DD-HH-MM.md. This is the
canonical template — test-reporter follows it verbatim (GitHub issue creation
flow lives in the test-reporter agent):
# Browser Test Report — {YYYY-MM-DD HH:MM}
- **Mode:** structured | exploratory
- **Base URL:** {baseURL}
- **Duration:** {Xm Ys}
- **Results:** {passed} passed / {failed} failed / {skipped} skipped
## Summary
| Route | Status | Severity | Notes |
|-------|--------|----------|-------|
| / | pass | — | |
## Failures
### {route} — {one-line description}
- **Severity:** critical | major | minor
- **Screenshot:** `screenshots/{slug}.png`
- **Repro steps:**
1. {step}
- **Observed:** {what happened}
- **Expected:** {what should happen}
## Warnings
- {skipped routes and non-critical observations, one bullet each}
Security: ALWAYS use AskUserQuestion before creating GitHub issues. Never auto-create. Warn that screenshots may contain sensitive data.