argusqa-os 9.7.4 → 9.7.5

package/README.md

@@ -4,7 +4,7 @@
 
 [![npm](https://img.shields.io/npm/v/argusqa-os?color=7C3AED)](https://www.npmjs.com/package/argusqa-os)
 [![MCP Server](https://glama.ai/mcp/servers/ironclawdevs27/Argus/badges/card.svg)](https://glama.ai/mcp/servers/ironclawdevs27/Argus)
-[![Harness](https://img.shields.io/badge/harness-688%2F688-4ADE80)](test-harness/)
+[![Harness](https://img.shields.io/badge/harness-738%2F738-4ADE80)](test-harness/)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
 
 **Argus catches the bugs your test suite misses — visual regressions, API loops, CSS drift, console noise, accessibility failures, and more — and delivers rich reports to Slack (or a local HTML dashboard).**
@@ -218,7 +218,7 @@ npm run report:pdf     # Export HTML report to A4 PDF (requires: npm install pup
 npm run server         # Start Slack slash-command server (port 3001)
 npm run init           # Interactive setup wizard
 npm run test:unit          # 61 unit tests — no Chrome required
-npm run test:harness       # 142-block correctness harness — requires Chrome
+npm run test:harness       # 144-block correctness harness — requires Chrome
 npm run test:harness:log   # same, but tees full output to harness-results.txt
 ```
 
@@ -342,7 +342,7 @@ Argus is a **complementary layer**, not a replacement for unit or E2E tests:
 
 ## Known Limitations
 
-All 688 harness assertions pass (`688/688`) — there are currently no known MCP- or Chrome-layer restrictions. Soft assertions (Lighthouse, performance traces) still require non-headless Chrome and are skipped in headless CI.
+All 738 harness assertions pass (`738/738`) — there are currently no known MCP- or Chrome-layer restrictions. Soft assertions (Lighthouse, performance traces) still require non-headless Chrome and are skipped in headless CI.
 
 ---
 
@@ -361,7 +361,7 @@ src/
     chrome-launcher.js  — npm run chrome / argus-chrome — launches Chrome with correct flags
     doctor.js           — npm run doctor / argus-doctor — pre-flight checks
     pr-validate.js      — headless CI entry point for GitHub Actions
-test-harness/           — 142-block correctness harness, 688 hard assertions, 62 fixture pages
+test-harness/           — 144-block correctness harness, 738 hard assertions, 60 fixture pages
 test/unit/              — 61 Vitest unit tests (no Chrome required)
 landing/                — Product landing page (React 19 + Vite + Tailwind)
 ```

package/glama.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://glama.ai/mcp/schemas/server.json",
   "name": "argus",
-  "description": "AI-powered QA harness that audits web apps via Chrome DevTools Protocol. Catches JS errors, network failures, a11y violations, SEO issues, security headers, CSS regressions, and more — directly from Claude conversations. 9 MCP tools: argus_audit (fast 8-analyzer pass), argus_audit_full (Lighthouse + memory + responsive), argus_compare (dev vs staging diff), argus_last_report (retrieve last JSON report), argus_watch_snapshot (live tab snapshot without navigating), argus_get_context (LLM-optimized context + fix loop with snapshot_id diff), argus_design_audit (Figma design fidelity — 13 finding types), argus_visual_diff (screenshot baseline comparison, updateBaseline flag), argus_pr_validate (PR diff → affected routes → targeted audit → blocked flag). Every finding is post-processed with intelligent baseline filtering (cross-run noise classifier) and root cause linking (recent git commits mapped to new findings). 142 test blocks, 688 hard assertions, 67 detection categories.",
+  "description": "AI-powered QA harness that audits web apps via Chrome DevTools Protocol. Catches JS errors, network failures, a11y violations, SEO issues, security headers, CSS regressions, and more — directly from Claude conversations. 9 MCP tools: argus_audit (fast 8-analyzer pass), argus_audit_full (Lighthouse + memory + responsive), argus_compare (dev vs staging diff), argus_last_report (retrieve last JSON report), argus_watch_snapshot (live tab snapshot without navigating), argus_get_context (LLM-optimized context + fix loop with snapshot_id diff), argus_design_audit (Figma design fidelity — 13 finding types), argus_visual_diff (screenshot baseline comparison, updateBaseline flag), argus_pr_validate (PR diff → affected routes → targeted audit → blocked flag). Every finding is post-processed with intelligent baseline filtering (cross-run noise classifier) and root cause linking (recent git commits mapped to new findings). 144 test blocks, 738 hard assertions, 67 detection categories.",
   "maintainers": ["ironclawdevs27"],
   "tools": [
     {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "argusqa-os",
-  "version": "9.7.4",
+  "version": "9.7.5",
   "mcpName": "io.github.ironclawdevs27/argus",
   "description": "Argus — AI-powered automated dev-testing platform using Chrome DevTools MCP and Claude Code",
   "keywords": [

package/src/adapters/browser.js

@@ -21,7 +21,24 @@ export class CdpBrowserAdapter {
   constructor(mcp) { this._mcp = mcp; }
 
   // ── Navigation ──────────────────────────────────────────────────────────────
-  navigate(url)            { return withRetry(() => this._mcp.navigate_page({ url }), { label: `navigate(${url})` }); }
+  // navigate_page reports failures as RESOLVED text ("Unable to navigate ...
+  // net::ERR_CONNECTION_REFUSED", "Could not connect to Chrome ..."), never as a
+  // thrown error. Unchecked, a dead target or dead browser produced a "clean"
+  // audit: analyzers ran against chrome-error://chromewebdata and emitted bogus
+  // findings (or none), and CI gates passed with Chrome down. Throw so failures
+  // propagate through the existing crawl error path.
+  navigate(url) {
+    return withRetry(async () => {
+      const resp = await this._mcp.navigate_page({ url });
+      if (typeof resp === 'string' &&
+          (resp.includes('Unable to navigate') ||
+           resp.includes('Could not connect to Chrome') ||
+           resp.includes('A dialog is open'))) {
+        throw new Error(`navigate(${url}) failed: ${resp.split('\n')[0].slice(0, 200)}`);
+      }
+      return resp;
+    }, { label: `navigate(${url})` });
+  }
 
   // ── Evaluation & snapshots ──────────────────────────────────────────────────
   evaluate(fn)             { return this._mcp.evaluate_script({ function: fn }); }
@@ -41,14 +58,52 @@ export class CdpBrowserAdapter {
   hover(uid)               { return this._mcp.hover({ uid }); }
   drag(src, tgt)           { return this._mcp.drag({ from_uid: src, to_uid: tgt }); }
   uploadFile(uid, filePath) { return this._mcp.upload_file({ uid, filePath }); }
-  handleDialog(accept, promptText = '') { return this._mcp.handle_dialog({ accept, promptText }); }
-  waitFor(opts)            { return this._mcp.wait_for(opts); }
+  // handle_dialog wire schema is { action: 'accept'|'dismiss', promptText? } — sending
+  // { accept: bool } is rejected by the tool's input validation (and the rejection comes
+  // back as a resolved error-text response, so the failure was silent in production).
+  handleDialog(accept, promptText = '') {
+    const args = { action: accept ? 'accept' : 'dismiss' };
+    if (promptText) args.promptText = promptText;
+    return this._mcp.handle_dialog(args);
+  }
+  // wait_for requires text as a non-empty string ARRAY. A bare string is rejected by
+  // input validation, and { state: 'networkidle' } is not part of the tool's schema at
+  // all — both shapes used to resolve to error text and silently wait for nothing.
+  waitFor(opts = {})       {
+    if (typeof opts.text === 'string') opts = { ...opts, text: [opts.text] };
+    if (opts.state === 'networkidle') return this.#waitForNetworkIdle();
+    return this._mcp.wait_for(opts);
+  }
+
+  // Bounded network-quiet poll: resolves once the page's resource-timing entry count
+  // is stable across two consecutive 250 ms polls, or after 3 s — whichever is first.
+  async #waitForNetworkIdle() {
+    let prev = -1;
+    for (let i = 0; i < 12; i++) {
+      const raw = await this.evaluate(`() => performance.getEntriesByType('resource').length`);
+      const count = Number(typeof raw === 'object' ? raw?.result ?? 0 : raw) || 0;
+      if (count === prev) return;
+      prev = count;
+      await new Promise(r => setTimeout(r, 250));
+    }
+  }
 
   // ── Viewport ────────────────────────────────────────────────────────────────
   emulate(viewport)              { return this._mcp.emulate({ viewport }); }
   emulateCpu(rate)               { return this._mcp.emulate({ cpuThrottlingRate: rate }); }
   emulateColorScheme(scheme)     { return this._mcp.emulate({ colorScheme: scheme }); }
-  emulateReducedMotion(pref)     { return this._mcp.emulate({ reducedMotion: pref }); }
+  // chrome-devtools-mcp@1.1.1's emulate tool has no reduced-motion capability — the
+  // unsupported argument comes back as RESOLVED error text ("Unknown argument"), not a
+  // thrown error, so callers' graceful-skip catch paths (motion-analyzer) never ran and
+  // analysis proceeded unemulated. Surface it as a real error; if a future upstream
+  // version adds the argument, the call succeeds and emulation lights up automatically.
+  async emulateReducedMotion(pref) {
+    const resp = await this._mcp.emulate({ reducedMotion: pref });
+    if (typeof resp === 'string' && resp.includes('Unknown argument')) {
+      throw new Error(`emulate does not support reducedMotion in this chrome-devtools-mcp version: ${resp.slice(0, 120)}`);
+    }
+    return resp;
+  }
   resize(w, h)                   { return this._mcp.resize_page({ width: w, height: h }); }
 
   // ── Network & performance ───────────────────────────────────────────────────

package/src/orchestration/orchestrator.js

@@ -370,54 +370,13 @@ function analyzeNetworkPerformance(perfEntries, pageUrl) {
   return bugs;
 }
 
-// ── Performance Budgets ────────────────────────────────────────────────────────
-
-async function checkPerformanceBudgets(browser, url) {
-  const violations = [];
-
-  try {
-    await browser.startTrace();
-    await new Promise(r => setTimeout(r, 3000));
-    const trace    = await browser.stopTrace();
-    const insights = await browser.analyzeInsight({ insightSetId: trace?.insightSetId ?? trace?.id ?? trace });
-
-    const metrics = insights?.metrics ?? insights?.performanceMetrics ?? {};
-
-    const checks = [
-      { key: 'LCP',  value: metrics.largestContentfulPaint ?? metrics.LCP,  budget: thresholds.perf.LCP,  unit: 'ms' },
-      { key: 'CLS',  value: metrics.cumulativeLayoutShift  ?? metrics.CLS,  budget: thresholds.perf.CLS,  unit: ''   },
-      { key: 'FID',  value: metrics.totalBlockingTime ?? metrics.TBT ?? metrics.FID, budget: thresholds.perf.FID, unit: 'ms' },
-      { key: 'TTFB', value: metrics.timeToFirstByte   ?? metrics.TTFB,      budget: thresholds.perf.TTFB, unit: 'ms' },
-    ];
-
-    for (const { key, value, budget, unit } of checks) {
-      if (value == null) continue;
-      if (value > budget) {
-        violations.push({
-          type:      'performance_budget',
-          metric:    key,
-          value:     `${value}${unit}`,
-          budget:    `${budget}${unit}`,
-          message:   `Performance budget exceeded: ${key} = ${value}${unit} (budget: ${budget}${unit})`,
-          severity:  'warning',
-          url,
-        });
-      }
-    }
-  } catch (err) {
-    logger.warn(`[ARGUS] Performance trace skipped for ${url}: ${err.message}`);
-  }
-
-  return violations;
-}
-
 // ── Cheap Crawl (called ×2 for flakiness detection) ───────────────────────────
 
 /**
  * Cheap detections for one route.
  * Runs: console, network, JS errors, blank page, API frequency, contracts,
  *       SEO, security, content, CSS, debugger statements, duplicate ids, screenshot.
- * Does NOT run: Lighthouse, perf budgets, network perf, redirect chain, broken links, cache headers.
+ * Does NOT run: Lighthouse, network perf, redirect chain, broken links, cache headers.
  */
 export async function crawlRouteCheap(route, baseUrl, mcp) {
   const browser = new CdpBrowserAdapter(mcp);
@@ -757,8 +716,11 @@ export async function crawlRouteCheap(route, baseUrl, mcp) {
 
 /**
  * Expensive/deterministic analyzers for one route — called ONCE per route.
- * Runs: network perf, redirect chain, perf budgets, Lighthouse,
+ * Runs: network perf, redirect chain, Lighthouse,
  *       broken internal links, cache headers.
+ * (Core Web Vitals — LCP/CLS/TTFB — are emitted by the web-vitals-analyzer
+ *  registerExpensive plugin, which is headless-compatible; the old trace-based
+ *  perf-budget path was removed as dead + redundant.)
  */
 export async function crawlRouteExpensive(route, baseUrl, mcp) {
   const browser = new CdpBrowserAdapter(mcp);
@@ -805,9 +767,6 @@ export async function crawlRouteExpensive(route, baseUrl, mcp) {
     logger.warn(`[ARGUS] Redirect chain check skipped for ${url}: ${err.message}`);
   }
 
-  // Performance budget check
-  errors.push(...(await checkPerformanceBudgets(browser, url)));
-
   // Full Lighthouse audit (capped at LIGHTHOUSE_TIMEOUT_MS to prevent indefinite hang)
   errors.push(...(await Promise.race([
     checkLighthouse(browser, url),

package/src/utils/mcp-client.js

@@ -181,11 +181,14 @@ export async function createMcpClient() {
         // MCP returns { content: [{ type, text|data }] } — extract the value
         const content = result?.content;
         if (Array.isArray(content) && content.length > 0) {
-          const item = content[0];
-          if (item.type === 'image') {
-            // take_screenshot returns base64 image data — return in a shape callers expect
-            return { data: item.data, mimeType: item.mimeType ?? 'image/png' };
+          // take_screenshot returns [text caption, image] — the image is NOT content[0],
+          // so scan the whole array for it. Reading only content[0] returned the caption
+          // string and starved every screenshot consumer of image data.
+          const img = content.find(c => c.type === 'image');
+          if (img) {
+            return { data: img.data, mimeType: img.mimeType ?? 'image/png' };
           }
+          const item = content[0];
           if (item.type === 'text') {
             const text = item.text;
             // chrome-devtools-mcp wraps evaluate_script results in a markdown code block: