form-tester 0.7.0 → 0.7.2

package/.claude/skills/form-tester/INSTRUCTIONS.md

@@ -10,17 +10,21 @@ Install skill files:
 form-tester install
 ```
 
-Run the CLI:
-```
-form-tester
-```
-
-Non-interactive mode (best for AI agents):
+Test modes:
 ```
+# AI mode (default) — no prompts:
 form-tester test <url> --auto
 form-tester test <url> --auto --pnr 12345 --persona ung-mann
+
+# Human mode — prompts for persona, scenario, person selection:
+form-tester test <url> --human
+
+# Full interactive CLI:
+form-tester
 ```
 
+When the user asks for human/interactive mode, use `--human`. Otherwise default to `--auto`.
+
 Commands:
 ```
 /setup
@@ -43,8 +47,9 @@ Notes:
 - Use the next-step checklist printed by the CLI (cookies, person selection, validation fix, Dokumenter verification, save HTML/PDF).
 - If the error modal appears on save or submit ("Det skjedde en feil under innsending av skjema. Prøv igjen senere."), open DevTools -> Network before retrying. Then try resubmitting once. If it persists, find the failed request and capture the Correlation ID header in test_results.txt.
 - Use `--help` or `-h` to print the command list without starting the prompt.
-- Playwright CLI commands are available from this skill when needed.
-- IMPORTANT: All screenshots taken during a test run MUST use `--full-page` to capture the entire page, not just the viewport. This applies to every screenshot command: `playwright-cli screenshot --filename "..." --full-page`
+- IMPORTANT: Always use `form-tester exec` instead of `playwright-cli` directly. This records all commands for replay. Same syntax: `form-tester exec fill e1 "value"`, `form-tester exec click e3`, `form-tester exec close` (finalizes recording).
+- Replay a previous run: `form-tester replay output/form-id/timestamp/recording.json`
+- IMPORTANT: All screenshots taken during a test run MUST use `--full-page` to capture the entire page, not just the viewport. This applies to every screenshot command: `form-tester exec screenshot --filename "..." --full-page`
 - IMPORTANT: Take a full-page screenshot EVERY TIME the page changes. This includes: after clicking any action button (Neste, Forrige, Send inn, etc.), after a step/page transition, after form validation errors appear, after modals open, and after submission. Name screenshots descriptively (e.g., step1_filled.png, step2_before_submit.png, submit_result.png).
 
 Test flow (when /test is triggered):
@@ -65,7 +70,7 @@ Dokumenter verification (only when modal confirms storage):
 2. The document list loads sorted newest first. The first entry should match the form title.
 3. Click "Se detaljer" on the first document, then click "Åpne dokumentet".
 4. IMPORTANT - document capture depends on format:
-   - PDF documents: DOWNLOAD the file. Use `playwright-cli pdf --filename "$OUTPUT_DIR/document.pdf"` or save via browser download. Do NOT screenshot PDFs.
-   - HTML documents: Take a FULL-PAGE screenshot of the ENTIRE document (`playwright-cli screenshot --filename "$OUTPUT_DIR/document_screenshot.png" --full-page`). HTML documents cannot be downloaded as files, so the full-page screenshot is the primary artifact. Also save the snapshot and raw HTML.
+   - PDF documents: DOWNLOAD the file. Use `form-tester exec pdf --filename "$OUTPUT_DIR/document.pdf"` or save via browser download. Do NOT screenshot PDFs.
+   - HTML documents: Take a FULL-PAGE screenshot of the ENTIRE document (`form-tester exec screenshot --filename "$OUTPUT_DIR/document_screenshot.png" --full-page`). HTML documents cannot be downloaded as files, so the full-page screenshot is the primary artifact. Also save the snapshot and raw HTML.
    - XML/other formats: Note the document type in test_results.txt and skip capture.
 5. Include the document verification results in test_results.txt (document title, whether it matched the form h1, document type: HTML/PDF/XML).

package/.claude/skills/form-tester/SKILL.md

@@ -20,7 +20,10 @@ form-tester install --global   # or install to ~/.claude/skills/
 form-tester test <url> --auto
 form-tester test <url> --auto --pnr 12345 --persona ung-mann --scenario "test validation"
 
-# Interactive CLI:
+# Interactive mode (prompts for persona, scenario, etc.):
+form-tester test <url> --human
+
+# Full interactive CLI:
 form-tester
 ```
 

package/.github/copilot-instructions.md

@@ -13,16 +13,21 @@ form-tester install
 ## Running the CLI
 
 ```bash
-# Interactive
-form-tester
-
-# Non-interactive (best for AI agents)
+# AI mode (default) — no prompts:
 form-tester test <url> --auto
 form-tester test <url> --auto --pnr 12345 --persona ung-mann --scenario "test validation"
+
+# Human mode — prompts for persona, scenario, person selection:
+form-tester test <url> --human
+
+# Full interactive CLI:
+form-tester
 ```
 
 Persona IDs: `ung-mann`, `gravid-kvinne`, `eldre-kvinne`, `kronisk-syk-mann`. Defaults to "noen" if omitted.
 
+When the user asks for `--human` mode, use that flag. Otherwise default to `--auto`.
+
 ## Commands
 
 | Command | Description |
@@ -36,17 +41,22 @@ Persona IDs: `ung-mann`, `gravid-kvinne`, `eldre-kvinne`, `kronisk-syk-mann`. De
 | `/clear` | Clear session |
 | `/quit` | Exit CLI |
 
-## Playwright CLI
+## Playwright CLI (use via form-tester exec)
+
+IMPORTANT: Always use `form-tester exec` instead of `playwright-cli` directly. This records all commands for replay.
 
-You can also use `playwright-cli` commands directly for browser automation:
+```bash
+form-tester exec open https://example.com
+form-tester exec snapshot
+form-tester exec fill e1 "value"
+form-tester exec click e3
+form-tester exec screenshot --filename=page.png --full-page
+form-tester exec close           # finalizes recording
+```
 
+Replay a previous run:
 ```bash
-playwright-cli open https://example.com
-playwright-cli snapshot
-playwright-cli fill e1 "value"
-playwright-cli click e3
-playwright-cli screenshot --filename=page.png --full-page
-playwright-cli close
+form-tester replay output/form-id/timestamp/recording.json
 ```
 
 ## Test Flow
@@ -73,6 +83,6 @@ After submission, read the modal text:
 - If it does NOT mention Dokumenter -> skip verification, note in test_results.txt.
 
 Document capture depends on format:
-- **PDF documents**: download the file (`playwright-cli pdf --filename "..." `). Do NOT screenshot PDFs.
-- **HTML documents**: take a full-page screenshot of the ENTIRE document (`playwright-cli screenshot --filename "..." --full-page`). HTML documents cannot be downloaded, so the screenshot is the primary artifact.
+- **PDF documents**: download the file (`form-tester exec pdf --filename "..." `). Do NOT screenshot PDFs.
+- **HTML documents**: take a full-page screenshot of the ENTIRE document (`form-tester exec screenshot --filename "..." --full-page`). HTML documents cannot be downloaded, so the screenshot is the primary artifact.
 - **XML/other**: note the type in test_results.txt and skip capture.

package/README.md

@@ -32,20 +32,41 @@ Edit `form-tester.config.json` and set your `pnr`.
 
 ## Usage
 
-### Default: non-interactive (`--auto`)
-
-AI agents (Claude Code, Copilot) will use `--auto` mode by default. No prompts, just runs the test:
+### Test modes
 
 ```bash
+# AI mode (default) — no prompts, uses defaults:
 form-tester test <url> --auto
 form-tester test <url> --auto --pnr 12345 --persona ung-mann --scenario "test validation"
+
+# Human mode — prompts for persona, scenario, person selection:
+form-tester test <url> --human
+
+# Full interactive CLI:
+form-tester
 ```
 
 Persona IDs: `ung-mann`, `gravid-kvinne`, `eldre-kvinne`, `kronisk-syk-mann`
 
+### Recording & Replay
+
+Every test run records all commands to `recording.json`. Use `form-tester exec` instead of `playwright-cli` to ensure all commands are recorded:
+
+```bash
+form-tester exec fill e1 "value"
+form-tester exec click e3
+form-tester exec screenshot --filename "page.png" --full-page
+form-tester exec close           # finalizes recording
+```
+
+Replay a previous run:
+```bash
+form-tester replay output/form-id/timestamp/recording.json
+```
+
 ### Interactive CLI
 
-For manual use without an AI agent:
+Full interactive CLI with all commands:
 
 ```bash
 form-tester

package/form-tester.js

@@ -6,7 +6,7 @@ const { spawn, execSync } = require("child_process");
 
 const CONFIG_PATH = path.join(process.cwd(), "form-tester.config.json");
 const OUTPUT_BASE = path.resolve(process.cwd(), "output");
-const LOCAL_VERSION = "0.7.0";
+const LOCAL_VERSION = "0.7.2";
 const RECOMMENDED_PERSON = "Uromantisk Direktør";
 
 // Recording — persisted to disk so `form-tester exec` can append across processes
@@ -1300,6 +1300,17 @@ async function main() {
     process.exit(0);
   }
 
+  if (args[0] === "test" && args.includes("--human")) {
+    const config = loadConfig();
+    const url = args.find((a) => a.startsWith("http"));
+    if (!url) {
+      console.error("Usage: form-tester test <url> --human");
+      process.exit(1);
+    }
+    await handleTest(url, config);
+    process.exit(0);
+  }
+
   if (args[0] === "test" && args.includes("--auto")) {
     const config = loadConfig();
     const url = args.find((a) => a.startsWith("http"));
@@ -1362,5 +1373,8 @@ module.exports = {
   getPersonaById,
   formatPersonaList,
   promptScenario,
+  startRecording,
+  appendToRecording,
+  finalizeRecording,
 };
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "form-tester",
-  "version": "0.7.0",
+  "version": "0.7.2",
   "description": "AI-powered form testing skill for /skjemautfyller forms using Playwright CLI. Works with Claude Code and GitHub Copilot.",
   "main": "form-tester.js",
   "bin": {

package/tests/form-tester.test.js

@@ -1,6 +1,10 @@
 const test = require("node:test");
 const assert = require("node:assert/strict");
 
+const fs = require("node:fs");
+const path = require("node:path");
+const os = require("node:os");
+
 const {
   extractFormId,
   extractPnrFromUrl,
@@ -14,6 +18,9 @@ const {
   getPersonaById,
   formatPersonaList,
   promptScenario,
+  startRecording,
+  appendToRecording,
+  finalizeRecording,
 } = require("../form-tester");
 
 test("extractFormId returns form id from skjemautfyller URL", () => {
@@ -118,3 +125,58 @@ test("each persona has a unique id", () => {
 test("promptScenario is exported as a function", () => {
   assert.equal(typeof promptScenario, "function");
 });
+
+// --- Recording tests ---
+
+function makeTmpDir() {
+  return fs.mkdtempSync(path.join(os.tmpdir(), "form-tester-test-"));
+}
+
+test("startRecording creates recording.json with empty commands", () => {
+  const dir = makeTmpDir();
+  const filePath = startRecording(dir);
+  assert.equal(filePath, path.join(dir, "recording.json"));
+  const data = JSON.parse(fs.readFileSync(filePath, "utf8"));
+  assert.equal(data.commandCount, 0);
+  assert.deepEqual(data.commands, []);
+  assert.ok(data.startedAt);
+  assert.equal(data.completedAt, null);
+  fs.rmSync(dir, { recursive: true });
+});
+
+test("appendToRecording adds commands to recording file", () => {
+  const dir = makeTmpDir();
+  const filePath = startRecording(dir);
+  appendToRecording(filePath, ["open", "https://example.com"]);
+  appendToRecording(filePath, ["fill", "e1", "hello"]);
+  appendToRecording(filePath, ["click", "e3"]);
+  const data = JSON.parse(fs.readFileSync(filePath, "utf8"));
+  assert.equal(data.commandCount, 3);
+  assert.deepEqual(data.commands[0].args, ["open", "https://example.com"]);
+  assert.deepEqual(data.commands[1].args, ["fill", "e1", "hello"]);
+  assert.deepEqual(data.commands[2].args, ["click", "e3"]);
+  assert.ok(data.commands[0].timestamp);
+  fs.rmSync(dir, { recursive: true });
+});
+
+test("finalizeRecording sets completedAt timestamp", () => {
+  const dir = makeTmpDir();
+  const filePath = startRecording(dir);
+  appendToRecording(filePath, ["snapshot"]);
+  const result = finalizeRecording(filePath);
+  assert.equal(result, filePath);
+  const data = JSON.parse(fs.readFileSync(filePath, "utf8"));
+  assert.ok(data.completedAt);
+  assert.equal(data.commandCount, 1);
+  fs.rmSync(dir, { recursive: true });
+});
+
+test("finalizeRecording returns null for missing file", () => {
+  assert.equal(finalizeRecording("/nonexistent/recording.json"), null);
+  assert.equal(finalizeRecording(null), null);
+});
+
+test("appendToRecording is resilient to missing file", () => {
+  // Should not throw
+  appendToRecording("/nonexistent/recording.json", ["click", "e1"]);
+});