E2E Agent Kit (Gemini/Gravity + VPS)

Production-oriented starter package for end-to-end web app testing with a Testsprite-style workflow.

What you get

• Structured test catalog (P0/P1)
• Real executable JSON test specs (Playwright-based runner)
• Deterministic run scripts (smoke/regression)
• Artifact folders (screenshots/logs/reports)
• Simple report portal (HTTP server) so you can revisit results anytime

Quick start

cd e2e-agent-kit
npm install
npm run setup:playwright
npm run bootstrap
cp .env.example .env
# edit .env with STAGING_URL and credentials
npm run smoke
npm run report:serve

Then open:

• http://<your-vps-ip>:8787/

Modes (anti-confusion)

• Platform mode (core): internal scaffold/demo validation
• Product mode (ag): real Anti Gravity validation

Main commands

• npm run test:platform — full daily gate in platform mode (core)
• npm run test:ag — full daily gate in product mode (ag)
• npm run smoke — run P0 smoke pack
• npm run regression — run P0+P1 regression
• npm run e2e:run — run by priority (PRIORITY=High,Medium npm run e2e:run)
• Multi-project runs: set PROJECT_KEY=<project> (e.g., PROJECT_KEY=ag npm run regression) to prefix run IDs for filtering
• Friendly project names in report explorer are configured in spec/projects.json
• npm run ai:generate-spec -- --story=e2e/stories/US-001-login-and-create.md --severity=P1 --id=P1-301 — generate draft JSON spec from user story
• npm run ai:refine-spec -- --spec=e2e/specs/p1/<file>.json --feedback="increase timeout and add wait for dashboard" — refine spec from natural language feedback
• npm run project:onboard -- --project=ag2 --name="Anti Gravity 2" --base-url=https://ag2.example.com — scaffold a new project onboarding pack
• npm run dashboard — generate consolidated dashboard snapshot (PROJECT_KEY=ag npm run dashboard for project-specific)
• npm run security — run security/session baseline test pack
• npm run mcp:daily-gate — run daily full gate (endpoints + API smoke + regression + security)
• npm run prod:status — evaluate latest readiness and print GO/NO-GO + reasons
• npm run spec:validate -- --run-id=<run-id> — compare executed tests with product requirements/spec mapping
• npm run spec:validate:combined — compare requirements against latest regression + security runs
• npm run deploy:frontend — one-command frontend deploy (uses dist/ by default)
• npm run check:endpoints — verify all public URLs + MCP auth guard
• npm run report:explorer — generate side-by-side report explorer page
• npm run report:serve — serve reports locally on port 8787

Optional: keep report portal always on (tmux)

tmux new -s e2e-reports -d 'cd /root/.openclaw/workspace/e2e-agent-kit && npm run report:serve'

Notes

• The current executor uses Playwright + JSON specs in e2e/specs/p0/*.json.
• Update selectors to your app's data-testid attributes.
• Save auth state in e2e/state/auth/ for stable repeated runs.
• Keep secrets in .env (git-ignored), never in committed files.
• Reports are generated per-run under reports/e2e/<run-id>/.
• Video recording is enabled per test case (reports/e2e/<run-id>/videos/).
• Step-by-step screenshots are auto-captured by default (reports/e2e/<run-id>/step-shots/). Set RECORD_STEP_SCREENSHOTS=false to disable.