greenrun-cli 0.2.12 → 0.2.14

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "greenrun-cli",
-  "version": "0.2.12",
+  "version": "0.2.14",
   "description": "CLI and MCP server for Greenrun - browser test management for Claude Code",
   "type": "module",
   "main": "dist/server.js",

package/templates/commands/procedures.md

@@ -118,9 +118,24 @@ Call this via `browser_run_code`. If `auth_mode` is `none`, skip this step.
 
 Gather all tests that have scripts (previously scripted + newly generated from Step 3).
 
-1. **Fetch and write test files**: For each scripted test, call `get_test(test_id)` to retrieve the full script content, then write it to `/tmp/greenrun-tests/{test_id}.spec.ts`. Fetch scripts in parallel to minimize latency.
+**0. Clean up** — run `rm -rf /tmp/greenrun-tests` via Bash to clear any stale files from a previous run.
 
-2. **Write config**: Write `/tmp/greenrun-tests/playwright.config.ts`:
+**1. Fetch scripts and write test files** — launch one Task agent per scripted test, all in parallel. These are just API calls + file writes so they don't conflict. Also write the Playwright config directly (it's small).
+
+For each scripted test, launch in parallel:
+
+```
+Task tool with:
+- subagent_type: "general-purpose"
+- max_turns: 5
+- model: "haiku"
+- prompt: "Fetch the Playwright script for test {test_id} and write it to a file.
+  1. Call `get_test(\"{test_id}\")` to fetch the test
+  2. Write the `script` field to `/tmp/greenrun-tests/{test_id}.spec.ts` using the Write tool
+  3. Return: \"{test_name} | written\""
+```
+
+While the agents run, write `/tmp/greenrun-tests/playwright.config.ts` directly:
 
 ```ts
 import { defineConfig } from '@playwright/test';
@@ -131,23 +146,24 @@ export default defineConfig({
   reporter: [['json', { outputFile: 'results.json' }]],
   use: {
     baseURL: '{base_url}',
-    storageState: '/tmp/greenrun-auth-state.json',  // omit 'use.storageState' entirely if auth_mode is 'none'
+    // include storageState ONLY if auth_mode is not 'none':
+    storageState: '/tmp/greenrun-auth-state.json',
   },
 });
 ```
 
-Replace `{base_url}` with the project's base_url.
+Wait for all agents to complete before executing.
 
-3. **Execute**: Run via Bash:
+**2. Execute** — run via Bash:
 ```
 npx playwright test --config /tmp/greenrun-tests/playwright.config.ts
 ```
 
-4. **Parse results**: Read `/tmp/greenrun-tests/results.json`. Map each result back to a run ID via the filename: `{test_id}.spec.ts` → test_id → find the matching run_id from the batch.
+**3. Parse results**: Read `/tmp/greenrun-tests/results.json`. Map each result back to a run ID via the filename: `{test_id}.spec.ts` → test_id → find the matching run_id from the batch.
 
-5. **Report results**: Call `complete_run(run_id, status, result_summary)` for each test. Map Playwright statuses: `passed` → `passed`, `failed`/`timedOut` → `failed`, other → `error`.
+**4. Report results**: Call `complete_run(run_id, status, result_summary)` for each test. Map Playwright statuses: `passed` → `passed`, `failed`/`timedOut` → `failed`, other → `error`.
 
-6. **Clean up**: Call `browser_close` to reset the MCP browser context.
+**5. Clean up**: Call `browser_close` to reset the MCP browser context.
 
 ### Step 6: Circuit breaker