debs-obrien Verify debbie.codes
Drive the live Nuxt 3 + Nuxt Content site at http://127.0.0.1:8000 through the real browser path. Prefer this skill’s CLI (control-debbie-codes.mjs), which wraps the repo’s existing @playwright/test / Playwright stack. Each drive command launches an ephemeral headless Chromium and closes it before exit (the Nuxt server stays up). Do not invent a second browser automation stack. Leave .agents/skills/playwright-cli and .agents/skills/add-content alone; they are separate Debbie workflows.
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs --help lists commands, flags, and examples.
Pick a flow
Interview facts for this repo:
| Concern | Finding |
|---|---|
| Surface | Public marketing / content site (home, about, blog, videos, podcasts, courses). Primary: web UI. |
| Run | npm run dev → Nuxt on 127.0.0.1:8000 (see README.md, nuxt.config.ts devServer, .cursor/environment.json). |
| Drive | Existing Playwright harness (playwright.config.ts, tests/*.spec.ts, CI .github/workflows/playwright.yml). This skill’s CLI reuses Playwright; optional ad-hoc: npx playwright test or .agents/skills/playwright-cli. |
| Observe | Screenshots + ARIA snapshots under .cursor/skills/verify-debbie-codes/artifacts/<run-id>/; HTTP identity; page URL/title. |
| Isolate | Default port 8000. One owned instance per state file in /tmp/debbie-codes-verify/state.json. Pass --port to launch another; pass --reuse to adopt an already-healthy server. Never kill a process this CLI did not start (owned: false). |
Launch
From the repo root:
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs launch --json
# or adopt whatever is already answering on :8000
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs launch --reuse --json
Ready when doctor reports healthy (HTTP 200 and page HTML contains Debbie). The CLI waits up to 120s for Nuxt. Logs for a owned launch: /tmp/debbie-codes-verify/<run-id>-nuxt.log.
Teardown is cleanup (below). Do not pkill -f nuxt / kill by process name.
Doctor
Run first whenever anything looks off:
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs doctor --json
Pass only when: base URL responds OK, identity string Debbie is present, and (if we own the process) the recorded pid is still alive. A failing doctor means relaunch or fix the environment before driving.
Drive
Prefer stable ARIA handles used by this repo’s Playwright specs: roles (heading, link, navigation, region, article), accessible names, and the blog search placeholder Search....
Low-level (compose a custom path):
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs goto /
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs expect --role heading --name "Debbie O'Brien"
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs click --role link --name Blog
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs fill --placeholder "Search..." --value playwright
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs expect --role heading --name "/Search Results/i"
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs expect-url "/\\/blog/"
Mapped feature recipes (one command; captures evidence):
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs drive home --json
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs drive navigation --json
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs drive blog --json
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs drive videos --json
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs drive about --json
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs drive tags-and-search --json
Read the matching file under features/ before claiming a feature is verified. For broader coverage, also run the existing suite: npx playwright test (same app URL; config reuses a healthy server when not in CI).
Hydration gotcha (from tests/blog.spec.ts / tests/blog-filters.spec.ts): cold loads can swallow the first nav click before the SPA router attaches. Retry click-until-URL when composing custom drives; the drive recipes already wait/toPass where needed.
Evidence
Proof standards:
- Exercise the real user path (nav links, listing → article, search box, tag chips) — not internal Nuxt APIs or content-collection queries as the primary proof.
- Capture the action and the resulting state (URL + visible heading/region), not only a final screenshot.
- Side effects here are mostly client-visible (filtered lists, tag pages). Confirm with ARIA snapshot + URL.
- No mocks: talk to the running site.
Artifacts live at:
.cursor/skills/verify-debbie-codes/artifacts/<run-id>/
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs screenshot --out .cursor/skills/verify-debbie-codes/artifacts/manual.png
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs snapshot --out .cursor/skills/verify-debbie-codes/artifacts/manual.aria.txt
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs evidence list --json
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs evidence path
drive <feature> writes *-proof.png, *-proof.aria.txt, and *-proof.json into the current run’s evidence directory.
Cleanup
# see what would stop (never destructive without confirmation via omitting --dry-run)
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs cleanup --dry-run --json
node .cursor/skills/verify-debbie-codes/control-debbie-codes.mjs cleanup --json
Cleanup stops the Nuxt process and Chromium session this CLI started (owned: true). It removes /tmp/debbie-codes-verify/state.json and ephemeral browser profile dirs under that temp folder. It never deletes .cursor/skills/verify-debbie-codes/artifacts/. After cleanup, confirm evidence still exists (evidence list or ls the path printed by cleanup).
If launch used --reuse on a server you did not start, cleanup will not kill that server.
Helpers
| Helper | Role |
|---|---|
control-debbie-codes.mjs |
Launch / doctor / drive / evidence / cleanup CLI (this skill). |
bin/verify |
Isolated launch on port 8010 with its own process-group cleanup. Use when you must not share the default Playwright port. |
npx playwright test |
Full regression suite already wired in CI. |
.agents/skills/playwright-cli |
Interactive snapshot/click exploration when you need a human-style browser session — not a replacement for this skill’s proof loop. |
Invocation from repo root only (paths are relative to the skill and resolve the repo root as three levels up from the skill file).
Feature map
The behavior inventory lives in features/. Each file uses the same four H2s: Sub-features, How to get to it (user POV), Driving it with control-debbie-codes, Gotchas.
Maintenance
When routes, nav labels, or search UI change, update the matching features/*.md and drive recipes in the same PR, or run /maintain-verification-skill if that pstack skill is available.