breadfast-qa

The canonical AI QA Companion — the workflow, not the story outputs.

• CLAUDE.md + docs/ai/** — the canonical workflow + knowledge base (single source of truth).
• qa-workflow/ — workflow definitions, skills, and the qa-cli.js gate tooling.
• .claude/skills/ — the Claude Code entrypoints (/qa-shift-left, /qa-validate, /qa-full).
• automation/ — shared page objects, helpers, config, the BrowserStack tooling (automation/browserstack/) and the legacy Playwright suite (automation/legacy/).

Everything here is self-contained: no folder outside this repository is required. The one external dependency is the Java/Appium framework used for automation generation, whose path is configurable (QA_FRAMEWORK_PATH → automation/config/framework.js) and which reports a clear "not found" rather than assuming a drive letter.

Per-story artifacts are not committed. A story folder (<TICKET>/) is created at the repo root during a run and is gitignored — requirements analysis, Figma captures, test cases, execution reports, screenshots, evidence and defects all stay local to whoever ran the story. Reusable tooling belongs in automation/; story data does not belong in git.

Quick start

npm install                       # @playwright/test, mysql2, ssh2, pdf-parse
npx playwright install chromium
cp automation/config/credentials.local.example.js automation/config/credentials.local.js
# fill in your own Jira token + BrowserStack login (or set env vars) — never commit it

Then run a workflow from Claude Code: /qa-shift-left, /qa-validate, or /qa-full.

Run the legacy Playwright suite:

npx playwright test --config=automation/legacy/playwright.config.js

History

The qa-platform/ app was removed on 2026-08-10 — a legacy execution engine, deferred since 2026-07-15 and superseded by Claude Code, with zero runtime coupling to this workflow. The design documents still cited by live docs were kept at docs/ai/architecture/legacy-qa-platform/; the app's own history remains on github.com/Breadfast/qa-platform.