@equinor/fusion-framework-cli-plugin-copilot 1.0.0

package/dist/esm/commands/app/login.js

@@ -0,0 +1,71 @@
+import { spawn } from 'node:child_process';
+import { existsSync, readFileSync } from 'node:fs';
+import { join } from 'node:path';
+import { ab, resolveAppKey, cleanup, resetDaemon, sleep, stripAnsi, waitForServer, pollConsole, } from '../../utils/index.js';
+/**
+ * Interactive login mode for the `copilot app` command.
+ *
+ * Starts the application dev server, opens a headed browser window for
+ * manual MSAL authentication, and polls the browser console for the Fusion
+ * `"Done"` message that signals successful initialisation. The browser is
+ * left open after login so the user can finish any manual steps before
+ * pressing Ctrl+C.
+ *
+ * @param absAppPath - Absolute path to the Fusion application directory
+ * @param port - Port number for the dev server
+ * @param host - Host address for the dev server
+ */
+export async function runLogin(absAppPath, port, host) {
+    const pkgJsonPath = join(absAppPath, 'package.json');
+    if (!existsSync(pkgJsonPath)) {
+        console.error(`❌ No package.json found at ${absAppPath}`);
+        process.exit(1);
+    }
+    const pkgJson = JSON.parse(readFileSync(pkgJsonPath, 'utf-8'));
+    if (!pkgJson.name) {
+        console.error(`❌ package.json at ${absAppPath} has no "name" field`);
+        process.exit(1);
+    }
+    let serverProcess;
+    console.log('📦 Resolving app manifest...');
+    const appKey = resolveAppKey(absAppPath);
+    console.log(`🚀 Starting server for ${appKey}...`);
+    serverProcess = spawn('ffc', ['app', 'dev', '--host', host, '--port', String(port)], {
+        cwd: absAppPath,
+        stdio: 'inherit',
+        detached: false,
+    });
+    const serverUrl = `http://localhost:${port}`;
+    const ready = await waitForServer(serverUrl, 60);
+    if (!ready) {
+        console.error('❌ Server failed to start within 60s');
+        cleanup(serverProcess);
+        process.exit(1);
+    }
+    const appUrl = `${serverUrl}/apps/${appKey}`;
+    console.log(`✅ Server ready at ${appUrl}`);
+    console.log('🌐 Opening headed browser — log in with your Equinor account...');
+    // Kill any stale daemon/Chrome so the new --headed --profile flags take effect
+    resetDaemon();
+    ab(['open', appUrl, '--headed'], 60_000);
+    // Ctrl+C handler: stop the dev server but leave the browser open
+    process.on('SIGINT', () => {
+        console.log('\n👋 Stopping server... (browser left open)');
+        cleanup(serverProcess);
+        process.exit(0);
+    });
+    // Wait for the Fusion app to initialise (browser console emits "✓ Done" after auth)
+    console.log('⏳ Waiting for app to initialize after login... (Ctrl+C to finish early)');
+    const initialized = await pollConsole((logs) => stripAnsi(logs).includes('Done'), 300_000);
+    if (initialized) {
+        await sleep(2000);
+        console.log('✅ Login complete — session saved.');
+    }
+    else {
+        console.log('⚠ Timed out waiting for app init. Session may still be saved.');
+    }
+    // Always wait for manual Ctrl+C so the user can finish what they're doing
+    console.log('🌐 Browser still open. Press Ctrl+C to exit.');
+    await new Promise(() => { }); // block until SIGINT
+}
+//# sourceMappingURL=login.js.map

package/dist/esm/commands/app/login.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"login.js","sourceRoot":"","sources":["../../../../src/commands/app/login.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,EAAqB,MAAM,oBAAoB,CAAC;AAC9D,OAAO,EAAE,UAAU,EAAE,YAAY,EAAE,MAAM,SAAS,CAAC;AACnD,OAAO,EAAE,IAAI,EAAE,MAAM,WAAW,CAAC;AAEjC,OAAO,EACL,EAAE,EACF,aAAa,EACb,OAAO,EACP,WAAW,EACX,KAAK,EACL,SAAS,EACT,aAAa,EACb,WAAW,GACZ,MAAM,sBAAsB,CAAC;AAE9B;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,KAAK,UAAU,QAAQ,CAAC,UAAkB,EAAE,IAAY,EAAE,IAAY;IAC3E,MAAM,WAAW,GAAG,IAAI,CAAC,UAAU,EAAE,cAAc,CAAC,CAAC;IACrD,IAAI,CAAC,UAAU,CAAC,WAAW,CAAC,EAAE,CAAC;QAC7B,OAAO,CAAC,KAAK,CAAC,8BAA8B,UAAU,EAAE,CAAC,CAAC;QAC1D,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IACD,MAAM,OAAO,GAAG,IAAI,CAAC,KAAK,CAAC,YAAY,CAAC,WAAW,EAAE,OAAO,CAAC,CAAsB,CAAC;IACpF,IAAI,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;QAClB,OAAO,CAAC,KAAK,CAAC,qBAAqB,UAAU,sBAAsB,CAAC,CAAC;QACrE,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,IAAI,aAAuC,CAAC;IAE5C,OAAO,CAAC,GAAG,CAAC,8BAA8B,CAAC,CAAC;IAC5C,MAAM,MAAM,GAAG,aAAa,CAAC,UAAU,CAAC,CAAC;IACzC,OAAO,CAAC,GAAG,CAAC,0BAA0B,MAAM,KAAK,CAAC,CAAC;IACnD,aAAa,GAAG,KAAK,CAAC,KAAK,EAAE,CAAC,KAAK,EAAE,KAAK,EAAE,QAAQ,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC,EAAE;QACnF,GAAG,EAAE,UAAU;QACf,KAAK,EAAE,SAAS;QAChB,QAAQ,EAAE,KAAK;KAChB,CAAC,CAAC;IAEH,MAAM,SAAS,GAAG,oBAAoB,IAAI,EAAE,CAAC;IAC7C,MAAM,KAAK,GAAG,MAAM,aAAa,CAAC,SAAS,EAAE,EAAE,CAAC,CAAC;IACjD,IAAI,CAAC,KAAK,EAAE,CAAC;QACX,OAAO,CAAC,KAAK,CAAC,qCAAqC,CAAC,CAAC;QACrD,OAAO,CAAC,aAAa,CAAC,CAAC;QACvB,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,MAAM,MAAM,GAAG,GAAG,SAAS,SAAS,MAAM,EAAE,CAAC;IAC7C,OAAO,CAAC,GAAG,CAAC,qBAAqB,MAAM,EAAE,CAAC,CAAC;IAC3C,OAAO,CAAC,GAAG,CAAC,iEAAiE,CAAC,CAAC;IAE/E,+EAA+E;IAC/E,WAAW,EAAE,CAAC;IACd,EAAE,CAAC,CAAC,MAAM,EAAE,MAAM,EAAE,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC;IAEzC,iEAAiE;IACjE,OAAO,CAAC,EAAE,CAAC,QAAQ,EAAE,GAAG,EAAE;QACxB,OAAO,CAAC,GAAG,CAAC,6CAA6C,CAAC,CAAC;QAC3D,OAAO,CAAC,aAAa,CAAC,CAAC;QACvB,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC,CAAC,CAAC;IAEH,oFAAoF;IACpF,OAAO,CAAC,GAAG,CAAC,yEAAyE,CAAC,CAAC;IACvF,MAAM,WAAW,GAAG,MAAM,WAAW,CACnC,CAAC,IAAY,EAAE,EAAE,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,EAClD,OAAO,CACR,CAAC;IAEF,IAAI,WAAW,EAAE,CAAC;QAChB,MAAM,KAAK,CAAC,IAAI,CAAC,CAAC;QAClB,OAAO,CAAC,GAAG,CAAC,mCAAmC,CAAC,CAAC;IACnD,CAAC;SAAM,CAAC;QACN,OAAO,CAAC,GAAG,CAAC,+DAA+D,CAAC,CAAC;IAC/E,CAAC;IAED,0EAA0E;IAC1E,OAAO,CAAC,GAAG,CAAC,8CAA8C,CAAC,CAAC;IAC5D,MAAM,IAAI,OAAO,CAAC,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC,CAAC,qBAAqB;AACpD,CAAC"}

package/dist/esm/commands/app/prompts/execute-step.prompt.js

@@ -0,0 +1,44 @@
+/**
+ * Creates the execution prompt for a single test step.
+ *
+ * The agent receives one plan step and executes its actions against the
+ * running application using browser tools, collecting evidence artifacts
+ * into the evidence directory.
+ *
+ * @param actions - The action strings to execute in order
+ * @param ctx - Runtime context with output directory and app URL
+ * @returns The fully interpolated execution prompt string
+ */
+export const createExecuteStepPrompt = (actions, ctx) => {
+    return `
+You are now EXECUTING one test step using real browser tools.
+
+ACTIONS: ${JSON.stringify(actions, null, 2)}
+
+APPLICATION URL: ${ctx.url}
+EVIDENCE DIRECTORY: ${ctx.outDir}/evidence
+
+Execute the "actions" array in order using the available browser tool calls.
+Collect rich evidence at every stage — especially on failure.
+
+Evidence rules (mandatory):
+- Before important actions: screenshot --annotate evidence/step-{criterion-slug}-before.png
+- After the step (always): screenshot --full --annotate evidence/step-{criterion-slug}-after.png
+- On any failure or suspicion:
+    errors > evidence/step-{criterion-slug}-errors.log
+    console > evidence/step-{criterion-slug}-console.log
+    snapshot -i --json > evidence/step-{criterion-slug}-snapshot.json
+- Use descriptive, unique filenames
+  Examples: step-login-button-missing-screenshot.png, step-dashboard-url.txt
+
+Important patterns:
+- Start most steps with: snapshot -i --annotate → use @eN refs for element stability
+- After navigation: wait --load networkidle or wait --url "**/pattern*"
+- On failure (non-zero exit, timeout, missing element, JS error): capture extra evidence
+- If flaky (timing, network): take 2–3 extra screenshots ~1 s apart
+
+When done, your work will be evaluated against the step's passEvidence and failEvidence.
+Do NOT output markdown or explanations — only use browser tool calls and collect evidence.
+  `.trim();
+};
+//# sourceMappingURL=execute-step.prompt.js.map

package/dist/esm/commands/app/prompts/execute-step.prompt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"execute-step.prompt.js","sourceRoot":"","sources":["../../../../../src/commands/app/prompts/execute-step.prompt.ts"],"names":[],"mappings":"AAEA;;;;;;;;;;GAUG;AACH,MAAM,CAAC,MAAM,uBAAuB,GAAG,CACrC,OAAiB,EACjB,GAA4B,EACpB,EAAE;IACV,OAAO;;;WAGE,IAAI,CAAC,SAAS,CAAC,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;;mBAExB,GAAG,CAAC,GAAG;sBACJ,GAAG,CAAC,MAAM;;;;;;;;;;;;;;;;;;;;;;;GAuB7B,CAAC,IAAI,EAAE,CAAC;AACX,CAAC,CAAC"}

package/dist/esm/commands/app/prompts/judge.prompt.js

@@ -0,0 +1,116 @@
+import { readdirSync, readFileSync, existsSync } from 'node:fs';
+import { join, extname } from 'node:path';
+/** Text file extensions that are safe to embed in the prompt. */
+const TEXT_EXTENSIONS = new Set(['.txt', '.json', '.jsonl', '.log', '.html', '.csv']);
+/**
+ * Reads `executions.jsonl` and all text evidence files from the run directory
+ * and returns them as a single string block suitable for prompt embedding.
+ *
+ * Binary files (screenshots) are listed by name but not inlined.
+ *
+ * @param outDir - The run output directory containing executions.jsonl and evidence/
+ * @returns Formatted string with execution records and text evidence
+ */
+function collectEvidence(outDir) {
+    const parts = [];
+    // ── executions.jsonl ──
+    const execPath = join(outDir, 'executions.jsonl');
+    if (existsSync(execPath)) {
+        parts.push('### executions.jsonl', readFileSync(execPath, 'utf-8').trim());
+    }
+    else {
+        parts.push('### executions.jsonl', '(file not found)');
+    }
+    // ── evidence/ directory ──
+    const evidenceDir = join(outDir, 'evidence');
+    if (existsSync(evidenceDir)) {
+        const files = readdirSync(evidenceDir).sort();
+        for (const file of files) {
+            const ext = extname(file).toLowerCase();
+            if (TEXT_EXTENSIONS.has(ext)) {
+                const content = readFileSync(join(evidenceDir, file), 'utf-8').trim();
+                parts.push(`### evidence/${file}`, content);
+            }
+            else {
+                parts.push(`### evidence/${file}`, `(binary file — ${ext} screenshot)`);
+            }
+        }
+    }
+    return parts.join('\n\n');
+}
+/**
+ * Creates the final-verdict prompt for the judge phase.
+ *
+ * The judge reviews ALL step executions and evidence holistically, then writes
+ * a single `verdict.json` with an overall pass/fail decision. This differs
+ * from the per-step evaluate prompt which assesses one criterion at a time.
+ *
+ * Execution records and text evidence files are embedded directly in the prompt
+ * so the model does not need file-reading tools or SDK attachment resolution.
+ *
+ * @param ctx - Runtime context containing the output directory and app URL
+ * @returns The fully interpolated judge prompt string
+ */
+export const createJudgePrompt = (ctx) => {
+    const evidence = collectEvidence(ctx.outDir);
+    return `
+All test steps have been executed. You are now the JUDGE.
+
+Review the evidence below and any screenshots using read_file, then produce a final verdict.
+
+<evidence>
+${evidence}
+</evidence>
+
+To inspect screenshots, use read_file with paths like "evidence/step-hello-fusion-header.jpg".
+Images are returned as inline vision content you can see.
+
+Save your verdict using write_file with path "verdict.json".
+You MUST call write_file. Do NOT output the JSON as a message.
+
+Evaluation rules — apply exactly:
+1. Auto-fail a criterion if:
+   - Any "Uncaught", TypeError, or ReferenceError appears in console/errors during that step
+   - The page shows an error, HTTP 500, white screen, or spinner lasting >15 s
+   - The expected element, text, or URL never appeared
+   - An explicit assertion (is visible, get text, get count) returned a failing result
+2. A criterion passes only when:
+   - All passEvidence conditions are clearly confirmed in screenshots (use read_file to view them), console, or URL
+   - No critical functional deviation is present
+3. Be lenient on:
+   - Console warnings, deprecations, analytics 404s
+   - Cosmetic shifts <5 px or layout differences that do not hide content
+   - Minor visual differences without functional impact
+
+Overall "pass" = true ONLY if every criterion is "ok": true.
+
+Output ONLY this JSON — no markdown fences, no prose:
+
+{
+  "pass": boolean,
+  "reasoning": "One-paragraph summary explaining the overall pass/fail outcome",
+  "steps": [
+    {
+      "criterion": "...",
+      "ok": true | false | "blocked" | "flaky",
+      "note": "Evidence-based verdict for this criterion"
+    }
+  ],
+  "ux": [
+    "Short UX observation or improvement suggestion based on what you saw in the screenshots and evidence"
+  ]
+}
+
+The "ux" array is for observations about the user experience that go beyond
+pass/fail — things a product owner or designer would want to know:
+- Accessibility issues (contrast, missing labels, keyboard traps)
+- Loading performance (slow renders, layout shifts, flash of unstyled content)
+- Visual polish (alignment, spacing, truncation, responsive issues)
+- Interaction hints (unclear affordances, missing hover states, confusing flow)
+- Content quality (typos, unclear copy, missing help text)
+Only include genuine observations from the evidence. Skip "ux" if nothing stands out.
+
+If evidence for a criterion is missing or ambiguous, fail it.
+  `.trim();
+};
+//# sourceMappingURL=judge.prompt.js.map

package/dist/esm/commands/app/prompts/judge.prompt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"judge.prompt.js","sourceRoot":"","sources":["../../../../../src/commands/app/prompts/judge.prompt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,YAAY,EAAE,UAAU,EAAE,MAAM,SAAS,CAAC;AAChE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAG1C,iEAAiE;AACjE,MAAM,eAAe,GAAG,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,OAAO,EAAE,QAAQ,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC;AAEtF;;;;;;;;GAQG;AACH,SAAS,eAAe,CAAC,MAAc;IACrC,MAAM,KAAK,GAAa,EAAE,CAAC;IAE3B,yBAAyB;IACzB,MAAM,QAAQ,GAAG,IAAI,CAAC,MAAM,EAAE,kBAAkB,CAAC,CAAC;IAClD,IAAI,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC;QACzB,KAAK,CAAC,IAAI,CAAC,sBAAsB,EAAE,YAAY,CAAC,QAAQ,EAAE,OAAO,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC;IAC7E,CAAC;SAAM,CAAC;QACN,KAAK,CAAC,IAAI,CAAC,sBAAsB,EAAE,kBAAkB,CAAC,CAAC;IACzD,CAAC;IAED,4BAA4B;IAC5B,MAAM,WAAW,GAAG,IAAI,CAAC,MAAM,EAAE,UAAU,CAAC,CAAC;IAC7C,IAAI,UAAU,CAAC,WAAW,CAAC,EAAE,CAAC;QAC5B,MAAM,KAAK,GAAG,WAAW,CAAC,WAAW,CAAC,CAAC,IAAI,EAAE,CAAC;QAC9C,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;YACzB,MAAM,GAAG,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,WAAW,EAAE,CAAC;YACxC,IAAI,eAAe,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC;gBAC7B,MAAM,OAAO,GAAG,YAAY,CAAC,IAAI,CAAC,WAAW,EAAE,IAAI,CAAC,EAAE,OAAO,CAAC,CAAC,IAAI,EAAE,CAAC;gBACtE,KAAK,CAAC,IAAI,CAAC,gBAAgB,IAAI,EAAE,EAAE,OAAO,CAAC,CAAC;YAC9C,CAAC;iBAAM,CAAC;gBACN,KAAK,CAAC,IAAI,CAAC,gBAAgB,IAAI,EAAE,EAAE,kBAAkB,GAAG,cAAc,CAAC,CAAC;YAC1E,CAAC;QACH,CAAC;IACH,CAAC;IAED,OAAO,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;AAC5B,CAAC;AAED;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,MAAM,iBAAiB,GAAG,CAAC,GAA4B,EAAU,EAAE;IACxE,MAAM,QAAQ,GAAG,eAAe,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;IAE7C,OAAO;;;;;;EAMP,QAAQ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAoDP,CAAC,IAAI,EAAE,CAAC;AACX,CAAC,CAAC"}

package/dist/esm/commands/app/prompts/plan.prompt.js

@@ -0,0 +1,47 @@
+/**
+ * Creates the planning prompt that turns a feature specification into
+ * a structured test plan.
+ *
+ * The agent produces a `plan.json` with one step per atomic acceptance
+ * criterion, each containing browser actions, pass evidence, and fail evidence.
+ *
+ * @param query - The feature specification / eval content to plan against
+ * @param ctx - Runtime context with output directory and app URL
+ * @returns The fully interpolated planning prompt string
+ */
+export const createPlanPrompt = (query, ctx) => {
+    return `
+Create a test plan for this feature specification:
+
+${query}
+
+Application URL: ${ctx.url}
+
+Save the plan using write_file with path "plan.json".
+
+JSON structure:
+
+{
+  "summary": "One-sentence description of what is being tested",
+  "steps": [
+    {
+      "scenario": "A logical user scenario (e.g. 'Open the landing page')",
+      "criteria": ["First thing to verify", "Second thing to verify"],
+      "pass": ["What pass looks like for first", "What pass looks like for second"],
+      "fail": ["What fail looks like for first", "What fail looks like for second"]
+    }
+  ]
+}
+
+Rules:
+- Group related criteria under one step when they share the same scenario
+  (e.g. same page, same state). Do NOT create separate steps for things
+  that can be verified in a single browser visit.
+- Only create multiple steps when there is a distinct user action between them
+  (navigation, click, form submission, etc.).
+- Do NOT include HOW to test — no actions, selectors, or tool calls.
+- passEvidence and failEvidence describe what a human would see.
+- Keep criteria directly traceable to the spec. Do NOT invent criteria.
+  `.trim();
+};
+//# sourceMappingURL=plan.prompt.js.map

package/dist/esm/commands/app/prompts/plan.prompt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"plan.prompt.js","sourceRoot":"","sources":["../../../../../src/commands/app/prompts/plan.prompt.ts"],"names":[],"mappings":"AAEA;;;;;;;;;;GAUG;AACH,MAAM,CAAC,MAAM,gBAAgB,GAAG,CAAC,KAAa,EAAE,GAA4B,EAAU,EAAE;IACtF,OAAO;;;EAGP,KAAK;;mBAEY,GAAG,CAAC,GAAG;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BvB,CAAC,IAAI,EAAE,CAAC;AACX,CAAC,CAAC"}

package/dist/esm/commands/app/prompts/step.prompt.js

@@ -0,0 +1,42 @@
+/**
+ * Creates the execution prompt for a single test step.
+ *
+ * The agent autonomously verifies all criteria for the scenario,
+ * takes evidence, and records results.
+ *
+ * @param step - The plan step with scenario and criteria
+ * @param ctx - Runtime context with output directory and app URL
+ * @returns The fully interpolated prompt string
+ */
+export const createStepPrompt = (step, ctx) => {
+    const slug = slugify(step.scenario);
+    const criteriaBlock = step.criteria
+        .map((c, i) => `${i + 1}. ${c}\n   Pass: ${step.pass[i] ?? ''}\n   Fail: ${step.fail[i] ?? ''}`)
+        .join('\n');
+    return `
+Scenario: ${step.scenario}
+
+Verify ALL of these criteria:
+${criteriaBlock}
+
+Use browser tools to verify. Take a screenshot when done:
+  browser_screenshot path="evidence/step-${slug}.jpg" annotate=true
+
+Then for EACH criterion, append ONE JSON line using append_file with path "executions.jsonl":
+{"criterion": "...", "ok": true|false, "note": "what you observed", "evidence": ["evidence/step-${slug}.jpg"]}
+  `.trim();
+};
+/**
+ * Converts text to a URL-safe slug for filenames.
+ *
+ * @param text - The text to slugify
+ * @returns Lowercase hyphenated string safe for filenames
+ */
+function slugify(text) {
+    return text
+        .toLowerCase()
+        .replace(/[^a-z0-9]+/g, '-')
+        .replace(/^-|-$/g, '')
+        .slice(0, 60);
+}
+//# sourceMappingURL=step.prompt.js.map

package/dist/esm/commands/app/prompts/step.prompt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"step.prompt.js","sourceRoot":"","sources":["../../../../../src/commands/app/prompts/step.prompt.ts"],"names":[],"mappings":"AAEA;;;;;;;;;GASG;AACH,MAAM,CAAC,MAAM,gBAAgB,GAAG,CAAC,IAAc,EAAE,GAA4B,EAAU,EAAE;IACvF,MAAM,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC;IACpC,MAAM,aAAa,GAAG,IAAI,CAAC,QAAQ;SAChC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,CAAC,GAAG,CAAC,KAAK,CAAC,cAAc,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,EAAE,cAAc,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC;SAC/F,IAAI,CAAC,IAAI,CAAC,CAAC;IAEd,OAAO;YACG,IAAI,CAAC,QAAQ;;;EAGvB,aAAa;;;2CAG4B,IAAI;;;kGAGmD,IAAI;GACnG,CAAC,IAAI,EAAE,CAAC;AACX,CAAC,CAAC;AAEF;;;;;GAKG;AACH,SAAS,OAAO,CAAC,IAAY;IAC3B,OAAO,IAAI;SACR,WAAW,EAAE;SACb,OAAO,CAAC,aAAa,EAAE,GAAG,CAAC;SAC3B,OAAO,CAAC,QAAQ,EAAE,EAAE,CAAC;SACrB,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;AAClB,CAAC"}

package/dist/esm/commands/app/prompts/system.dev-server.prompt.js

@@ -0,0 +1,21 @@
+/**
+ * Creates the Fusion dev-server–specific system prompt supplement.
+ *
+ * Describes the DOM structure of a Fusion Framework application served
+ * by the local dev server / dev-portal so the agent knows where to
+ * find app content in the page.
+ *
+ * @param ctx - Runtime context with the application URL
+ * @returns The dev-server DOM context block to append to the system prompt
+ */
+export const createDevServerContext = (ctx) => {
+    return `
+Fusion dev-server DOM:
+- App content lives inside <section id="app-section" style="display:contents">.
+- While loading, #app-section contains an EquinorLoader spinner — wait for it to disappear.
+- Once loaded, the app renders in an anonymous <div> child of #app-section.
+- Target #app-section for assertions, not <body>.
+- URL pattern: /apps/{appKey} → ${ctx.url}
+  `.trim();
+};
+//# sourceMappingURL=system.dev-server.prompt.js.map

package/dist/esm/commands/app/prompts/system.dev-server.prompt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"system.dev-server.prompt.js","sourceRoot":"","sources":["../../../../../src/commands/app/prompts/system.dev-server.prompt.ts"],"names":[],"mappings":"AAEA;;;;;;;;;GASG;AACH,MAAM,CAAC,MAAM,sBAAsB,GAAG,CAAC,GAA4B,EAAU,EAAE;IAC7E,OAAO;;;;;;kCAMyB,GAAG,CAAC,GAAG;GACtC,CAAC,IAAI,EAAE,CAAC;AACX,CAAC,CAAC"}

package/dist/esm/commands/app/prompts/system.prompt.js

@@ -0,0 +1,77 @@
+/**
+ * Creates the system prompt that sets the agent's identity and ground rules
+ * for the entire eval session.
+ *
+ * Injected once as the session-level system message. Individual phase prompts
+ * (plan, execute, evaluate, judge) are sent as user messages within this session.
+ *
+ * @param ctx - Runtime context with the application URL and output directory
+ * @returns The fully interpolated system prompt string
+ */
+export const createSystemPrompt = (ctx) => {
+    return `
+You are an expert software test agent with perfect knowledge of the feature specification,
+the application under test, and the testing tools at your disposal.
+
+Application under test: ${ctx.url}
+
+## File tools use RELATIVE paths
+
+The write_file, append_file, and read_file tools are scoped to an output directory.
+Always use **relative paths** like "plan.json", "executions.jsonl", "evidence/screenshot.jpg".
+Never use absolute paths with these tools.
+
+You will be guided through phases in order:
+1. PLAN    — produce a structured test plan (plan.json)
+2. EXECUTE — run each step's browser actions and collect evidence
+3. JUDGE   — review all evidence and deliver a final pass/fail verdict (verdict.json)
+
+All phases run in this single session. You have access to both browser tools
+and file tools throughout.
+
+## Tool inventory
+
+### File tools
+- **write_file** — write a file to the artifacts directory (plan.json, verdict.json)
+- **append_file** — append a line to a file (executions.jsonl)
+- **read_file** — read a file or list a directory from the artifacts directory.
+  Image files (png, jpg) are returned as inline vision content you can inspect.
+  Use this in the JUDGE phase to view screenshots taken during execution.
+
+### Browser tools
+- **browser_navigate** — open a URL
+- **browser_snapshot** — capture an accessibility snapshot (DOM tree). Use \`-i --json\` for interactive-only, JSON format.
+- **browser_screenshot** — capture a screenshot. Returns the image inline so you can see it.
+  Use \`--annotate --full\` for labeled full-page screenshots.
+- **browser_get_styles** — get computed styles of an element
+- **browser_eval** — evaluate JavaScript in the page context
+- **browser_click** — click an element (use @eN refs from snapshot, or CSS/role selectors)
+- **browser_fill** — fill an input field
+- **browser_type** — type text character by character
+- **browser_press_key** — press a keyboard key
+- **browser_hover** — hover over an element
+- **browser_select** — select an option from a dropdown
+- **browser_scroll** — scroll the page or an element
+- **browser_wait** — wait for a condition: \`networkidle\`, text, selector, or time (ms)
+- **browser_find** — find elements by role, text, label, placeholder, testid, or CSS
+- **browser_errors** — get console errors and uncaught exceptions
+- **browser_get_url** — get the current URL
+- **browser_go_back** — navigate back
+- **browser_reload** — reload the current page
+
+## Tool use rules
+- Act immediately. Do not deliberate or plan between tool calls — execute then evaluate.
+- Keep tool calls minimal. Prefer fewer, targeted calls over exhaustive probing.
+- After navigation, always \`browser_wait\` for \`networkidle\` before interacting.
+- Use \`@eN\` element refs from \`browser_snapshot\` for clicks, fills, and assertions.
+- Take ONE screenshot per step (after the action). Skip redundant before-screenshots.
+- Capture \`browser_errors\` only when something looks wrong, not on every step.
+- Do NOT use browser tools during the PLAN phase — only describe actions as strings.
+- Do NOT use browser tools during the JUDGE phase — use \`read_file\` to view evidence.
+
+## General rules
+- Save artifacts using write_file / append_file with RELATIVE paths (e.g. "plan.json").
+- Never fabricate evidence. If a tool call fails, report the failure honestly.
+  `.trim();
+};
+//# sourceMappingURL=system.prompt.js.map

package/dist/esm/commands/app/prompts/system.prompt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"system.prompt.js","sourceRoot":"","sources":["../../../../../src/commands/app/prompts/system.prompt.ts"],"names":[],"mappings":"AAEA;;;;;;;;;GASG;AACH,MAAM,CAAC,MAAM,kBAAkB,GAAG,CAAC,GAA4B,EAAU,EAAE;IACzE,OAAO;;;;0BAIiB,GAAG,CAAC,GAAG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2D9B,CAAC,IAAI,EAAE,CAAC;AACX,CAAC,CAAC"}

package/dist/esm/commands/app/server.js

@@ -0,0 +1,83 @@
+import { spawn, execSync } from 'node:child_process';
+import { existsSync, readFileSync } from 'node:fs';
+import { join } from 'node:path';
+import { resolveAppKey, cleanup, waitForServer } from '../../utils/index.js';
+/**
+ * Kills any process listening on the given port.
+ *
+ * Uses `lsof` to find PIDs bound to the port and sends SIGTERM.
+ * Failures are silently ignored (port may already be free).
+ *
+ * @param port - The TCP port to free
+ */
+function freePort(port) {
+    try {
+        const pids = execSync(`lsof -ti tcp:${port}`, { encoding: 'utf-8' })
+            .trim()
+            .split('\n')
+            .filter(Boolean);
+        for (const pid of pids) {
+            try {
+                process.kill(parseInt(pid, 10), 'SIGTERM');
+            }
+            catch {
+                // already dead
+            }
+        }
+        if (pids.length) {
+            console.log(`🔄 Killed stale process(es) on port ${port}: ${pids.join(', ')}`);
+        }
+    }
+    catch {
+        // lsof exits non-zero when no matches — port is free
+    }
+}
+/**
+ * Starts the Fusion app dev server for the given application path,
+ * or returns the caller-provided URL when the server is already running.
+ *
+ * @param absAppPath - Absolute path to the Fusion application directory
+ * @param options - Server configuration options
+ * @param options.port - Port number for the dev server
+ * @param options.host - Host address for the dev server
+ * @param options.url - When provided, skips server startup and uses this URL directly
+ * @param options.verbose - When true, inherits stdio from the spawned server process
+ * @returns The spawned server process (if any) and the resolved app URL
+ */
+export async function startAppServer(absAppPath, options) {
+    // Caller already has a running server — skip startup entirely
+    if (options.url) {
+        return { serverProcess: undefined, appUrl: options.url };
+    }
+    const pkgJsonPath = join(absAppPath, 'package.json');
+    if (!existsSync(pkgJsonPath)) {
+        console.error(`❌ No package.json found at ${absAppPath}`);
+        process.exit(1);
+    }
+    const pkgJson = JSON.parse(readFileSync(pkgJsonPath, 'utf-8'));
+    if (!pkgJson.name) {
+        console.error(`❌ package.json at ${absAppPath} has no "name" field`);
+        process.exit(1);
+    }
+    console.log('📦 Resolving app manifest...');
+    const appKey = resolveAppKey(absAppPath);
+    console.log(`🚀 Starting app server (${appKey})...`);
+    // Kill stale dev servers on the target port from previous crashed runs
+    freePort(options.port);
+    const serverProcess = spawn('ffc', ['app', 'serve', '--host', options.host, '--port', String(options.port)], {
+        cwd: absAppPath,
+        stdio: options.verbose ? 'inherit' : 'ignore',
+        detached: false,
+    });
+    const serverUrl = `http://localhost:${options.port}`;
+    const ready = await waitForServer(serverUrl, 60);
+    if (!ready) {
+        console.error('❌ Server failed to start within 60s');
+        cleanup(serverProcess);
+        process.exit(1);
+    }
+    const appUrl = `${serverUrl}/apps/${appKey}`;
+    console.log(`✅ Server ready at ${appUrl}`);
+    return { serverProcess, appUrl };
+}
+//# sourceMappingURL=server.js.map

package/dist/esm/commands/app/server.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"server.js","sourceRoot":"","sources":["../../../../src/commands/app/server.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,EAAE,QAAQ,EAAqB,MAAM,oBAAoB,CAAC;AACxE,OAAO,EAAE,UAAU,EAAE,YAAY,EAAE,MAAM,SAAS,CAAC;AACnD,OAAO,EAAE,IAAI,EAAE,MAAM,WAAW,CAAC;AAEjC,OAAO,EAAE,aAAa,EAAE,OAAO,EAAE,aAAa,EAAE,MAAM,sBAAsB,CAAC;AAU7E;;;;;;;GAOG;AACH,SAAS,QAAQ,CAAC,IAAY;IAC5B,IAAI,CAAC;QACH,MAAM,IAAI,GAAG,QAAQ,CAAC,gBAAgB,IAAI,EAAE,EAAE,EAAE,QAAQ,EAAE,OAAO,EAAE,CAAC;aACjE,IAAI,EAAE;aACN,KAAK,CAAC,IAAI,CAAC;aACX,MAAM,CAAC,OAAO,CAAC,CAAC;QACnB,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;YACvB,IAAI,CAAC;gBACH,OAAO,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,EAAE,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC;YAC7C,CAAC;YAAC,MAAM,CAAC;gBACP,eAAe;YACjB,CAAC;QACH,CAAC;QACD,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,OAAO,CAAC,GAAG,CAAC,uCAAuC,IAAI,KAAK,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACjF,CAAC;IACH,CAAC;IAAC,MAAM,CAAC;QACP,qDAAqD;IACvD,CAAC;AACH,CAAC;AAED;;;;;;;;;;;GAWG;AACH,MAAM,CAAC,KAAK,UAAU,cAAc,CAClC,UAAkB,EAClB,OAAuE;IAEvE,8DAA8D;IAC9D,IAAI,OAAO,CAAC,GAAG,EAAE,CAAC;QAChB,OAAO,EAAE,aAAa,EAAE,SAAS,EAAE,MAAM,EAAE,OAAO,CAAC,GAAG,EAAE,CAAC;IAC3D,CAAC;IAED,MAAM,WAAW,GAAG,IAAI,CAAC,UAAU,EAAE,cAAc,CAAC,CAAC;IACrD,IAAI,CAAC,UAAU,CAAC,WAAW,CAAC,EAAE,CAAC;QAC7B,OAAO,CAAC,KAAK,CAAC,8BAA8B,UAAU,EAAE,CAAC,CAAC;QAC1D,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IACD,MAAM,OAAO,GAAG,IAAI,CAAC,KAAK,CAAC,YAAY,CAAC,WAAW,EAAE,OAAO,CAAC,CAAsB,CAAC;IACpF,IAAI,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;QAClB,OAAO,CAAC,KAAK,CAAC,qBAAqB,UAAU,sBAAsB,CAAC,CAAC;QACrE,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,OAAO,CAAC,GAAG,CAAC,8BAA8B,CAAC,CAAC;IAC5C,MAAM,MAAM,GAAG,aAAa,CAAC,UAAU,CAAC,CAAC;IACzC,OAAO,CAAC,GAAG,CAAC,2BAA2B,MAAM,MAAM,CAAC,CAAC;IAErD,uEAAuE;IACvE,QAAQ,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IAEvB,MAAM,aAAa,GAAG,KAAK,CACzB,KAAK,EACL,CAAC,KAAK,EAAE,OAAO,EAAE,QAAQ,EAAE,OAAO,CAAC,IAAI,EAAE,QAAQ,EAAE,MAAM,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,EACxE;QACE,GAAG,EAAE,UAAU;QACf,KAAK,EAAE,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,QAAQ;QAC7C,QAAQ,EAAE,KAAK;KAChB,CACF,CAAC;IAEF,MAAM,SAAS,GAAG,oBAAoB,OAAO,CAAC,IAAI,EAAE,CAAC;IACrD,MAAM,KAAK,GAAG,MAAM,aAAa,CAAC,SAAS,EAAE,EAAE,CAAC,CAAC;IACjD,IAAI,CAAC,KAAK,EAAE,CAAC;QACX,OAAO,CAAC,KAAK,CAAC,qCAAqC,CAAC,CAAC;QACrD,OAAO,CAAC,aAAa,CAAC,CAAC;QACvB,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,MAAM,MAAM,GAAG,GAAG,SAAS,SAAS,MAAM,EAAE,CAAC;IAC7C,OAAO,CAAC,GAAG,CAAC,qBAAqB,MAAM,EAAE,CAAC,CAAC;IAE3C,OAAO,EAAE,aAAa,EAAE,MAAM,EAAE,CAAC;AACnC,CAAC"}