fimo 2.4.0-staging.8 → 2.4.0

package/assets/skills/fimo-cli/SKILL.md

@@ -91,6 +91,7 @@ Load the matching reference when the user's request fires its trigger. Recipes t
 - "Deploy / publish / go live / push my changes" → `references/deploy.md`
 - "Add a redirect / security headers / CSP / HSTS / clean URLs / trailing slash" → `references/hosting.md`
 - "Get the preview URL / open the preview / what's the URL for this Branch / refresh the preview link / I just want to see the site / the preview URL stopped working" → `references/preview.md`
+- "Check the site in a browser / is my site accessible / a11y or WCAG check / measure LCP or Web Vitals / why is the page slow / inspect network requests or console errors / Lighthouse score" → `references/site-checks.md`
 - "Schedule a command / cron / nightly publish / recurring checks / `fimo schedules`" → `references/schedules.md`
 - "Set env vars / configure API keys / add project secrets / local `.env.local` / `fimo vars` / `fimo secrets`" → `references/project-vars-and-secrets.md`
 - "React to form submissions / publish changes / agent runs / analytics reports / media activity / send events to Resend, Slack, GitHub, Zapier, or an HTTP endpoint / configure or debug outbound webhooks" → `references/webhooks.md`

package/assets/skills/fimo-cli/references/site-checks.md

@@ -0,0 +1,73 @@
+# Site checks
+
+Start from a URL the agent can actually reach, then use the inspection tools
+available in its runtime. This reference owns Fimo target selection and the
+meaning of the evidence. Browser and audit tools own their commands.
+
+## 1. Pick a target and name it
+
+- An already-running local development server.
+- An existing hosted Fimo preview: run `fimo preview url --json` and use
+  `data.sandbox_url` (see `references/preview.md`).
+- A URL the user supplied, including an existing published site.
+
+Never run `fimo deploy` just to retrieve a preview URL. Do not run `fimo
+publish` or `fimo deploy --publish` merely to create a measurement target.
+For any other deployment or publishing decision, follow the current run's
+authority rules and `references/deploy.md`.
+
+A local server or preview can prove how that target renders and behaves. Its
+timings describe that environment, not the published site. A supplied public
+URL can describe the live version, but one synthetic run is not field data.
+
+## 2. Use the tools that exist
+
+Inspect the available tools and skills before choosing a method.
+
+- If `agent-browser` is available, load its skill and version-matched `core`
+  guidance before using it.
+- Otherwise use equivalent browser automation, accessibility, HTTP, network,
+  profiling, or performance tools exposed by the current runtime.
+- HTTP-only tools can prove fetchability, status, headers, and raw content.
+  They cannot prove visual layout, interactions, or accessibility.
+- Fimo CLI does not produce Lighthouse reports, and `agent-browser` Web Vitals
+  are not Lighthouse scores. If the environment provides Lighthouse, use its
+  own guidance and name the target. If it does not, report that limitation and
+  use only measurements that can actually run.
+
+## 3. Pick the lens the question needs
+
+| The user asks                                                   | Evidence to obtain                            |
+| --------------------------------------------------------------- | --------------------------------------------- |
+| Does the page render, is the copy right, is an image broken     | browser snapshot and screenshot               |
+| Does this flow work, does the form submit, is the console clean | browser interaction and console output        |
+| Is it accessible, WCAG, contrast, labels, alt text              | accessibility audit                           |
+| Why does it feel slow, LCP, CLS, INP, TTFB, hydration           | Web Vitals and, when needed, a profiler trace |
+| Wrong data, a failing request, a slow API, response headers     | network log, HAR, or HTTP response            |
+| Does the page expose its own tools to agents                    | WebMCP page tools, when supported             |
+| What is its Lighthouse score                                    | an actual Lighthouse run                      |
+
+Run only the lenses the question needs. Two lenses with clear evidence beat a
+sweep of all seven.
+
+## 4. Report honestly
+
+- State the target URL and its kind (hosted preview, dev server, supplied URL),
+  the viewport, and when the checks ran.
+- Give the observed values with their units and how many runs produced them.
+  One run is one run; say so.
+- Keep what you observed separate from what you inferred, and list the lenses
+  you did not run.
+- Never claim a check that did not run. If the environment blocked it (no
+  browser, an expired sandbox, a target that would not load), report what
+  blocked it.
+
+For visitor behavior on the published site (pageviews, top pages, sources),
+use `references/analytics.md`. It reports real traffic rather than one browser
+run.
+
+## See also
+
+- `references/preview.md` for preview destinations and sandbox URL expiry.
+- `references/deploy.md` for preview deployment and publishing authorization.
+- `references/analytics.md` for published-site traffic.

package/dist/cli/bundle.json

@@ -1,8 +1,8 @@
 {
   "bundled": true,
   "bundler": "esbuild",
-  "bundledAt": "2026-09-07T09:32:43.178Z",
-  "cliVersion": "2.3.2",
+  "bundledAt": "2026-09-07T17:20:21.505Z",
+  "cliVersion": "2.4.0",
   "external": [
     "oxc-parser",
     "fsevents",