@salesforce/sfnext-audit

Deterministic pre-launch (go-live) readiness auditor for storefront-next customer projects, shipped as a TanStack Intent agent skill.

It captures Lighthouse against the deployed site, runs a zero-dependency static engine over the project, and produces a self-contained HTML dashboard + markdown report graded against the Salesforce standard launch bar — a GO / GO-WITH-FIXES / NO-GO verdict across performance, best practices, accessibility, SEO, security, and customization drift.

What you get

Every run writes, under .site-audit/:

• index.html — a self-contained dashboard (embeds the data, opens offline via file://). One card per captured page (home / PLP / PDP), a per-category breakdown, and the SSR render timeline. This is "what we saw" — share it directly.
• SITE-AUDIT-REPORT.md — the markdown report.
• audit.json — the deterministic single source of truth (see skills/site-audit/schema/).

Consumer quick start

You use this through a coding agent (Claude Code, Cursor, Copilot, Codex, …) that speaks TanStack Intent. See the Intent consumer guide.

1. Install into the storefront-next project you want to audit:

   pnpm add -D @salesforce/sfnext-audit   # or: npm install -D @salesforce/sfnext-audit

2. Confirm it's discovered:

   npx @tanstack/intent@latest list
   # → @salesforce/sfnext-audit  local  …  site-audit [lifecycle]

3. (Optional) wire skill-loading guidance into your agent config (AGENTS.md / CLAUDE.md / …):

   npx @tanstack/intent@latest install

4. Ask your agent for a pre-launch readiness audit of the project (give it the deployed/staging URL). The agent loads the skill —

   npx @tanstack/intent@latest load @salesforce/sfnext-audit#site-audit

   — then follows it: discovers the site's home / PLP / PDP pages, runs the engine, interprets the results per category, and reaches a verdict. The engine itself is available as a CLI:

   npx sfnext-audit --project <path> \
     --url home=<root> --url plp=<listing-url> --url pdp=<product-url> \
     --out <path>/.site-audit

5. Open .site-audit/index.html to see the dashboard.

A deployed-URL capture adds ~3–5 min per URL; a home+PLP+PDP run is ~9–15 min. Run npx sfnext-audit --help for all options (--lighthouse run|reuse|skip, --only <dims>, …).

Requirements

• Node.js ≥ 24
• A deployed/staging URL for the site (Lighthouse capture). Without one, performance degrades to static-only and the gap is reported.
• The engine analyzes the codebase statically — the project does not need to be runnable and you do not need its secrets.

Maintaining this package

The skill lives entirely under skills/site-audit/ (SKILL.md + the frozen bin/audit.mjs engine + lib/ + references/ + schema/ + audit.config.mjs). files: ["skills", "!skills/_artifacts"] in package.json is what gets published; the engine's imports resolve relative to their own location, so the whole tree ships together and runs from any consumer's node_modules.

npm install                                   # dev deps (@tanstack/intent)
npx @tanstack/intent@latest validate          # validate SKILL.md frontmatter + limits
npm pack --dry-run                            # preview the published tarball

Discovery relies on the intent field + the skills/ directory in package.json. Skill addressing is @salesforce/sfnext-audit#site-audit.

License

Apache-2.0.