Skip to content

Playwright is a framework for Web Testing and Automation. It allows testing Chromium, Firefox and WebKit with a single API.

chore: refactor init-agents (#37877)

Pavel Feldmandeca3acca8e2
diff --git a/examples/todomvc/.claude/agents/playwright-test-generator.md b/examples/todomvc/.claude/agents/playwright-test-generator.md
index b82e9bc3a1..276b2d737f 100644
--- a/examples/todomvc/.claude/agents/playwright-test-generator.md
+++ b/examples/todomvc/.claude/agents/playwright-test-generator.md
@@ -1,7 +1,7 @@
---
name: playwright-test-generator
description: Use this agent when you need to create automated browser tests using Playwright. Examples: <example>Context: User wants to test a login flow on their web application. user: 'I need a test that logs into my app at localhost:3000 with username [email protected] and password 123456, then verifies the dashboard page loads' assistant: 'I'll use the generator agent to create and validate this login test for you' <commentary> The user needs a specific browser automation test created, which is exactly what the generator agent is designed for. </commentary></example><example>Context: User has built a new checkout flow and wants to ensure it works correctly. user: 'Can you create a test that adds items to cart, proceeds to checkout, fills in payment details, and confirms the order?' assistant: 'I'll use the generator agent to build a comprehensive checkout flow test' <commentary> This is a complex user journey that needs to be automated and tested, perfect for the generator agent. </commentary></example>
-tools: Glob, Grep, Read, mcp__playwright-test__browser_click, mcp__playwright-test__browser_drag, mcp__playwright-test__browser_evaluate, mcp__playwright-test__browser_file_upload, mcp__playwright-test__browser_handle_dialog, mcp__playwright-test__browser_hover, mcp__playwright-test__browser_navigate, mcp__playwright-test__browser_press_key, mcp__playwright-test__browser_select_option, mcp__playwright-test__browser_snapshot, mcp__playwright-test__browser_type, mcp__playwright-test__browser_verify_element_visible, mcp__playwright-test__browser_verify_list_visible, mcp__playwright-test__browser_verify_text_visible, mcp__playwright-test__browser_verify_value, mcp__playwright-test__browser_wait_for, mcp__playwright-test__generator_read_log, mcp__playwright-test__generator_setup_page, mcp__playwright-test__generator_write_test
+tools: Read, Glob, Grep, mcp__playwright-test__browser_click, mcp__playwright-test__browser_drag, mcp__playwright-test__browser_evaluate, mcp__playwright-test__browser_file_upload, mcp__playwright-test__browser_handle_dialog, mcp__playwright-test__browser_hover, mcp__playwright-test__browser_navigate, mcp__playwright-test__browser_press_key, mcp__playwright-test__browser_select_option, mcp__playwright-test__browser_snapshot, mcp__playwright-test__browser_type, mcp__playwright-test__browser_verify_element_visible, mcp__playwright-test__browser_verify_list_visible, mcp__playwright-test__browser_verify_text_visible, mcp__playwright-test__browser_verify_value, mcp__playwright-test__browser_wait_for, mcp__playwright-test__generator_read_log, mcp__playwright-test__generator_setup_page, mcp__playwright-test__generator_write_test
model: sonnet
color: blue
---
diff --git a/examples/todomvc/.claude/agents/playwright-test-healer.md b/examples/todomvc/.claude/agents/playwright-test-healer.md
index 0a31a92a52..18cd979974 100644
--- a/examples/todomvc/.claude/agents/playwright-test-healer.md
+++ b/examples/todomvc/.claude/agents/playwright-test-healer.md
@@ -1,7 +1,7 @@
---
name: playwright-test-healer
description: Use this agent when you need to debug and fix failing Playwright tests. Examples: <example>Context: A developer has a failing Playwright test that needs to be debugged and fixed. user: 'The login test is failing, can you fix it?' assistant: 'I'll use the healer agent to debug and fix the failing login test.' <commentary> The user has identified a specific failing test that needs debugging and fixing, which is exactly what the healer agent is designed for. </commentary></example><example>Context: After running a test suite, several tests are reported as failing. user: 'Test user-registration.spec.ts is broken after the recent changes' assistant: 'Let me use the healer agent to investigate and fix the user-registration test.' <commentary> A specific test file is failing and needs debugging, which requires the systematic approach of the playwright-test-healer agent. </commentary></example>
-tools: Glob, Grep, Read, Write, Edit, MultiEdit, mcp__playwright-test__browser_console_messages, mcp__playwright-test__browser_evaluate, mcp__playwright-test__browser_generate_locator, mcp__playwright-test__browser_network_requests, mcp__playwright-test__browser_snapshot, mcp__playwright-test__test_debug, mcp__playwright-test__test_list, mcp__playwright-test__test_run
+tools: Read, Glob, Grep, Write, Edit, MultiEdit, mcp__playwright-test__browser_console_messages, mcp__playwright-test__browser_evaluate, mcp__playwright-test__browser_generate_locator, mcp__playwright-test__browser_network_requests, mcp__playwright-test__browser_snapshot, mcp__playwright-test__test_debug, mcp__playwright-test__test_list, mcp__playwright-test__test_run
model: sonnet
color: red
---
diff --git a/examples/todomvc/.claude/agents/playwright-test-planner.md b/examples/todomvc/.claude/agents/playwright-test-planner.md
index 283a88d7fe..afca58af3a 100644
--- a/examples/todomvc/.claude/agents/playwright-test-planner.md
+++ b/examples/todomvc/.claude/agents/playwright-test-planner.md
@@ -1,7 +1,7 @@
---
name: playwright-test-planner
description: Use this agent when you need to create comprehensive test plan for a web application or website. Examples: <example>Context: User wants to test a new e-commerce checkout flow. user: 'I need test scenarios for our new checkout process at https://mystore.com/checkout' assistant: 'I'll use the planner agent to navigate to your checkout page and create comprehensive test scenarios.' <commentary> The user needs test planning for a specific web page, so use the planner agent to explore and create test scenarios. </commentary></example><example>Context: User has deployed a new feature and wants thorough testing coverage. user: 'Can you help me test our new user dashboard at https://app.example.com/dashboard?' assistant: 'I'll launch the planner agent to explore your dashboard and develop detailed test scenarios.' <commentary> This requires web exploration and test scenario creation, perfect for the planner agent. </commentary></example>
-tools: Glob, Grep, Read, Write, mcp__playwright-test__browser_click, mcp__playwright-test__browser_close, mcp__playwright-test__browser_console_messages, mcp__playwright-test__browser_drag, mcp__playwright-test__browser_evaluate, mcp__playwright-test__browser_file_upload, mcp__playwright-test__browser_handle_dialog, mcp__playwright-test__browser_hover, mcp__playwright-test__browser_navigate, mcp__playwright-test__browser_navigate_back, mcp__playwright-test__browser_network_requests, mcp__playwright-test__browser_press_key, mcp__playwright-test__browser_select_option, mcp__playwright-test__browser_snapshot, mcp__playwright-test__browser_take_screenshot, mcp__playwright-test__browser_type, mcp__playwright-test__browser_wait_for, mcp__playwright-test__planner_setup_page
+tools: Read, Glob, Grep, Write, mcp__playwright-test__browser_click, mcp__playwright-test__browser_close, mcp__playwright-test__browser_console_messages, mcp__playwright-test__browser_drag, mcp__playwright-test__browser_evaluate, mcp__playwright-test__browser_file_upload, mcp__playwright-test__browser_handle_dialog, mcp__playwright-test__browser_hover, mcp__playwright-test__browser_navigate, mcp__playwright-test__browser_navigate_back, mcp__playwright-test__browser_network_requests, mcp__playwright-test__browser_press_key, mcp__playwright-test__browser_select_option, mcp__playwright-test__browser_snapshot, mcp__playwright-test__browser_take_screenshot, mcp__playwright-test__browser_type, mcp__playwright-test__browser_wait_for, mcp__playwright-test__planner_setup_page
model: sonnet
color: green
---
diff --git a/examples/todomvc/prompt.claudecode.md b/examples/todomvc/prompt.claudecode.md
deleted file mode 100644
index 21ee6db859..0000000000
--- a/examples/todomvc/prompt.claudecode.md
+++ /dev/null
@@ -1,14 +0,0 @@
-## Objective
-Test basic functionality of todo app.
-
-## Test setup: tests/template.spec.ts:11
-
-## Steps
-- Use `playwright-test-planner` subagent to create a test plan for "Basic operations".
- Use seed test from `tests/seed.spec.ts` to init page. Save the test plan as `specs/basic-operations.md`.
-
-- For each scenario in `specs/basic-operations.md`, use `playwright-test-generator`
- subagent to perform the scenario and generate the test source code into `tests/` folder.
- Process all scenarios sequentially, do not run in parallel.
-
-- Use `playwright-test-healer` subagent to fix the failing tests.
diff --git a/packages/playwright/src/agents/generateAgents.ts b/packages/playwright/src/agents/generateAgents.ts
index b349ad9234..6754573756 100644
--- a/packages/playwright/src/agents/generateAgents.ts
+++ b/packages/playwright/src/agents/generateAgents.ts
@@ -31,8 +31,8 @@ interface AgentHeader {
tools: string[];
}
-
interface Agent {
+ source: string;
header: AgentHeader;
instructions: string;
examples: string[];
@@ -41,11 +41,16 @@ interface Agent {
/* eslint-disable no-console */
class AgentParser {
+ static async loadAgents(): Promise<Agent[]> {
+ const files = await fs.promises.readdir(__dirname);
+ return Promise.all(files.filter(file => file.endsWith('.md')).map(file => AgentParser.parseFile(path.join(__dirname, file))));
+ }
+
static async parseFile(filePath: string): Promise<Agent> {
- const rawMarkdown = await fs.promises.readFile(filePath, 'utf-8');
- const { header, content } = this.extractYamlAndContent(rawMarkdown);
+ const source = await fs.promises.readFile(filePath, 'utf-8');
+ const { header, content } = this.extractYamlAndContent(source);
const { instructions, examples } = this.extractInstructionsAndExamples(content);
- return { header, instructions, examples };
+ return { source, header, instructions, examples };
}
static extractYamlAndContent(markdown: string): { header: AgentHeader; content: string } {
@@ -102,95 +107,241 @@ class AgentParser {
}
}
-const claudeToolMap = new Map<string, string[]>([
- ['ls', ['Glob']],
- ['grep', ['Grep']],
- ['read', ['Read']],
- ['edit', ['Edit', 'MultiEdit']],
- ['write', ['Write']],
-]);
-
-// Common MCP server configurations
-const commonMcpServers = {
- playwrightTest: {
- type: 'local',
- command: 'npx',
- args: ['playwright', 'run-test-mcp-server']
- }
-};
-
-function saveAsClaudeCode(agent: Agent): string {
- function asClaudeTool(tool: string): string {
- const [first, second] = tool.split('/');
- if (!second)
- return (claudeToolMap.get(first) || [first]).join(', ');
- return `mcp__${first}__${second}`;
+export class ClaudeGenerator {
+ static async init(config: FullConfigInternal, projectName: string) {
+ await initRepo(config, projectName);
+
+ const agents = await AgentParser.loadAgents();
+
+ await fs.promises.mkdir('.claude/agents', { recursive: true });
+ for (const agent of agents)
+ await writeFile(`.claude/agents/${agent.header.name}.md`, ClaudeGenerator.agentSpec(agent), 'πŸ€–', 'agent definition');
+
+ await writeFile('.mcp.json', JSON.stringify({
+ mcpServers: {
+ 'playwright-test': {
+ command: 'npx',
+ args: ['playwright', 'run-test-mcp-server'],
+ }
+ }
+ }, null, 2), 'πŸ”§', 'mcp configuration');
+
+ initRepoDone();
}
- const lines: string[] = [];
- lines.push(`---`);
- lines.push(`name: playwright-test-${agent.header.name}`);
- lines.push(`description: ${agent.header.description}. Examples: ${agent.examples.map(example => `<example>${example}</example>`).join('')}`);
- lines.push(`tools: ${agent.header.tools.map(tool => asClaudeTool(tool)).join(', ')}`);
- lines.push(`model: ${agent.header.model}`);
- lines.push(`color: ${agent.header.color}`);
- lines.push(`---`);
- lines.push('');
- lines.push(agent.instructions);
- return lines.join('\n');
+ static agentSpec(agent: Agent): string {
+ const claudeToolMap = new Map<string, string[]>([
+ ['search', ['Glob', 'Grep']],
+ ['read', ['Read']],
+ ['edit', ['Edit', 'MultiEdit']],
+ ['write', ['Write']],
+ ]);
+
+ function asClaudeTool(tool: string): string {
+ const [first, second] = tool.split('/');
+ if (!second)
+ return (claudeToolMap.get(first) || [first]).join(', ');
+ return `mcp__${first}__${second}`;
+ }
+
+ const lines: string[] = [];
+ lines.push(`---`);
+ lines.push(`name: ${agent.header.name}`);
+ lines.push(`description: ${agent.header.description}. Examples: ${agent.examples.map(example => `<example>${example}</example>`).join('')}`);
+ lines.push(`tools: ${agent.header.tools.map(tool => asClaudeTool(tool)).join(', ')}`);
+ lines.push(`model: ${agent.header.model}`);
+ lines.push(`color: ${agent.header.color}`);
+ lines.push(`---`);
+ lines.push('');
+ lines.push(agent.instructions);
+ return lines.join('\n');
+ }
}
-const opencodeToolMap = new Map<string, string[]>([
- ['ls', ['ls', 'glob']],
- ['grep', ['grep']],
- ['read', ['read']],
- ['edit', ['edit']],
- ['write', ['write']],
-]);
-
-function saveAsOpencodeJson(agents: Agent[]): string {
- function asOpencodeTool(tools: Record<string, boolean>, tool: string): void {
- const [first, second] = tool.split('/');
- if (!second) {
- for (const tool of opencodeToolMap.get(first) || [first])
- tools[tool] = true;
- } else {
- tools[`${first}*${second}`] = true;
+export class OpencodeGenerator {
+ static async init(config: FullConfigInternal, projectName: string) {
+ await initRepo(config, projectName);
+
+ const agents = await AgentParser.loadAgents();
+
+ await fs.promises.mkdir('.opencode/prompts', { recursive: true });
+ for (const agent of agents) {
+ const prompt = [agent.instructions];
+ prompt.push('');
+ prompt.push(...agent.examples.map(example => `<example>${example}</example>`));
+ await writeFile(`.opencode/prompts/${agent.header.name}.md`, prompt.join('\n'), 'πŸ€–', 'agent definition');
}
+ await writeFile('opencode.json', OpencodeGenerator.configuration(agents), 'πŸ”§', 'opencode configuration');
+
+ initRepoDone();
}
- const result: Record<string, any> = {};
- result['$schema'] = 'https://opencode.ai/config.json';
- result['mcp'] = {};
- result['tools'] = {
- 'playwright*': false,
- };
- result['agent'] = {};
- for (const agent of agents) {
- const tools: Record<string, boolean> = {};
- result['agent']['playwright-test-' + agent.header.name] = {
- description: agent.header.description,
- mode: 'subagent',
- prompt: `{file:.opencode/prompts/playwright-test-${agent.header.name}.md}`,
- tools,
+ static configuration(agents: Agent[]): string {
+ const opencodeToolMap = new Map<string, string[]>([
+ ['search', ['ls', 'glob', 'grep']],
+ ['read', ['read']],
+ ['edit', ['edit']],
+ ['write', ['write']],
+ ]);
+
+ const asOpencodeTool = (tools: Record<string, boolean>, tool: string) => {
+ const [first, second] = tool.split('/');
+ if (!second) {
+ for (const tool of opencodeToolMap.get(first) || [first])
+ tools[tool] = true;
+ } else {
+ tools[`${first}*${second}`] = true;
+ }
+ };
+
+ const result: Record<string, any> = {};
+ result['$schema'] = 'https://opencode.ai/config.json';
+ result['mcp'] = {};
+ result['tools'] = {
+ 'playwright*': false,
};
- for (const tool of agent.header.tools)
- asOpencodeTool(tools, tool);
+ result['agent'] = {};
+ for (const agent of agents) {
+ const tools: Record<string, boolean> = {};
+ result['agent'][agent.header.name] = {
+ description: agent.header.description,
+ mode: 'subagent',
+ prompt: `{file:.opencode/prompts/${agent.header.name}.md}`,
+ tools,
+ };
+ for (const tool of agent.header.tools)
+ asOpencodeTool(tools, tool);
+ }
+
+ result['mcp']['playwright-test'] = {
+ type: 'local',
+ command: ['npx', 'playwright', 'run-test-mcp-server'],
+ enabled: true,
+ };
+
+ return JSON.stringify(result, null, 2);
}
+}
+export class AgentGenerator {
+ static async init(config: FullConfigInternal, projectName: string) {
+ const agentsFolder = process.env.AGENTS_FOLDER;
+ if (!agentsFolder) {
+ console.error('AGENTS_FOLDER environment variable is not set');
+ return;
+ }
- const server = commonMcpServers.playwrightTest;
- result['mcp']['playwright-test'] = {
- type: server.type,
- command: [server.command, ...server.args],
- enabled: true,
- };
+ await initRepo(config, projectName);
+
+ const agents = await AgentParser.loadAgents();
+
+ await fs.promises.mkdir(agentsFolder, { recursive: true });
+ for (const agent of agents)
+ await writeFile(`${agentsFolder}/${agent.header.name}.md`, agent.source, 'πŸ€–', 'agent definition');
+
+ console.log('πŸ”§ MCP configuration');
+ console.log(JSON.stringify({
+ mcpServers: {
+ 'playwright-test': {
+ type: 'stdio',
+ command: 'npx',
+ args: [
+ `--prefix=${path.resolve(process.cwd())}`,
+ 'playwright',
+ 'run-test-mcp-server',
+ `--headless`,
+ `--config=${path.resolve(process.cwd())}`,
+ ],
+ tools: ['*']
+ }
+ }
+ }, null, 2));
- return JSON.stringify(result, null, 2);
+ initRepoDone();
+ }
}
-async function loadAgents(): Promise<Agent[]> {
- const files = await fs.promises.readdir(__dirname);
- return Promise.all(files.filter(file => file.endsWith('.md')).map(file => AgentParser.parseFile(path.join(__dirname, file))));
+export class VSCodeGenerator {
+ static async init(config: FullConfigInternal, projectName: string) {
+ await initRepo(config, projectName);
+ const agents = await AgentParser.loadAgents();
+
+ const nameMap = new Map<string, string>([
+ ['playwright-test-planner', ' 🎭 planner'],
+ ['playwright-test-generator', '🎭 generator'],
+ ['playwright-test-healer', '🎭 healer'],
+ ]);
+
+ await fs.promises.mkdir('.github/chatmodes', { recursive: true });
+ for (const agent of agents)
+ await writeFile(`.github/chatmodes/${nameMap.get(agent.header.name)}.chatmode.md`, VSCodeGenerator.agentSpec(agent), 'πŸ€–', 'chatmode definition');
+
+ await fs.promises.mkdir('.vscode', { recursive: true });
+
+ const mcpJsonPath = '.vscode/mcp.json';
+ let mcpJson: any = {
+ servers: {},
+ inputs: []
+ };
+ try {
+ mcpJson = JSON.parse(fs.readFileSync(mcpJsonPath, 'utf8'));
+ } catch {
+ }
+
+ if (!mcpJson.servers)
+ mcpJson.servers = {};
+
+ mcpJson.servers['playwright-test'] = {
+ type: 'stdio',
+ command: 'npx',
+ args: ['playwright', 'run-test-mcp-server'],
+ cwd: '${workspaceFolder}',
+ };
+ await writeFile(mcpJsonPath, JSON.stringify(mcpJson, null, 2), 'πŸ”§', 'mcp configuration');
+
+ initRepoDone();
+ }
+
+ static agentSpec(agent: Agent): string {
+ const vscodeToolMap = new Map<string, string[]>([
+ ['search', ['search/listDirectory', 'search/fileSearch', 'search/textSearch']],
+ ['read', ['search/readFile']],
+ ['edit', ['edit/editFiles']],
+ ['write', ['edit/createFile', 'edit/createDirectory']],
+ ]);
+ const vscodeToolsOrder = ['edit/createFile', 'edit/createDirectory', 'edit/editFiles', 'search/fileSearch', 'search/textSearch', 'search/listDirectory', 'search/readFile'];
+ const vscodeMcpName = 'playwright-test';
+
+ function asVscodeTool(tool: string): string | string[] {
+ const [first, second] = tool.split('/');
+ if (second)
+ return `${vscodeMcpName}/${second}`;
+ return vscodeToolMap.get(first) || first;
+ }
+ const tools = agent.header.tools.map(asVscodeTool).flat().sort((a, b) => {
+ // VSCode insists on the specific tools order when editing agent config.
+ const indexA = vscodeToolsOrder.indexOf(a);
+ const indexB = vscodeToolsOrder.indexOf(b);
+ if (indexA === -1 && indexB === -1)
+ return a.localeCompare(b);
+ if (indexA === -1)
+ return 1;
+ if (indexB === -1)
+ return -1;
+ return indexA - indexB;
+ }).map(tool => `'${tool}'`).join(', ');
+
+ const lines: string[] = [];
+ lines.push(`---`);
+ lines.push(`description: ${agent.header.description}.`);
+ lines.push(`tools: [${tools}]`);
+ lines.push(`---`);
+ lines.push('');
+ lines.push(agent.instructions);
+ for (const example of agent.examples)
+ lines.push(`<example>${example}</example>`);
+
+ return lines.join('\n');
+ }
}
async function writeFile(filePath: string, content: string, icon: string, description: string) {
@@ -234,120 +385,6 @@ function initRepoDone() {
console.log('βœ… Done.');
}
-export async function initClaudeCodeRepo(config: FullConfigInternal, projectName: string) {
- await initRepo(config, projectName);
-
- const agents = await loadAgents();
-
- await fs.promises.mkdir('.claude/agents', { recursive: true });
- for (const agent of agents)
- await writeFile(`.claude/agents/playwright-test-${agent.header.name}.md`, saveAsClaudeCode(agent), 'πŸ€–', 'agent definition');
-
- await writeFile('.mcp.json', JSON.stringify({
- mcpServers: {
- 'playwright-test': {
- command: commonMcpServers.playwrightTest.command,
- args: commonMcpServers.playwrightTest.args,
- }
- }
- }, null, 2), 'πŸ”§', 'mcp configuration');
-
- initRepoDone();
-}
-
-const vscodeToolMap = new Map<string, string[]>([
- ['ls', ['search/listDirectory', 'search/fileSearch']],
- ['grep', ['search/textSearch']],
- ['read', ['search/readFile']],
- ['edit', ['edit/editFiles']],
- ['write', ['edit/createFile', 'edit/createDirectory']],
-]);
-const vscodeToolsOrder = ['edit/createFile', 'edit/createDirectory', 'edit/editFiles', 'search/fileSearch', 'search/textSearch', 'search/listDirectory', 'search/readFile'];
-const vscodeMcpName = 'playwright-test';
-function saveAsVSCodeChatmode(agent: Agent): string {
- function asVscodeTool(tool: string): string | string[] {
- const [first, second] = tool.split('/');
- if (second)
- return `${vscodeMcpName}/${second}`;
- return vscodeToolMap.get(first) || first;
- }
- const tools = agent.header.tools.map(asVscodeTool).flat().sort((a, b) => {
- // VSCode insists on the specific tools order when editing agent config.
- const indexA = vscodeToolsOrder.indexOf(a);
- const indexB = vscodeToolsOrder.indexOf(b);
- if (indexA === -1 && indexB === -1)
- return a.localeCompare(b);
- if (indexA === -1)
- return 1;
- if (indexB === -1)
- return -1;
- return indexA - indexB;
- }).map(tool => `'${tool}'`).join(', ');
-
- const lines: string[] = [];
- lines.push(`---`);
- lines.push(`description: ${agent.header.description}.`);
- lines.push(`tools: [${tools}]`);
- lines.push(`---`);
- lines.push('');
- lines.push(agent.instructions);
- for (const example of agent.examples)
- lines.push(`<example>${example}</example>`);
-
- return lines.join('\n');
-}
-
-export async function initVSCodeRepo(config: FullConfigInternal, projectName: string) {
- await initRepo(config, projectName);
- const agents = await loadAgents();
-
- await fs.promises.mkdir('.github/chatmodes', { recursive: true });
- for (const agent of agents)
- await writeFile(`.github/chatmodes/${agent.header.name === 'planner' ? ' ' : ''}🎭 ${agent.header.name}.chatmode.md`, saveAsVSCodeChatmode(agent), 'πŸ€–', 'chatmode definition');
-
- await fs.promises.mkdir('.vscode', { recursive: true });
-
- const mcpJsonPath = '.vscode/mcp.json';
- let mcpJson: any = {
- servers: {},
- inputs: []
- };
- try {
- mcpJson = JSON.parse(fs.readFileSync(mcpJsonPath, 'utf8'));
- } catch {
- }
-
- if (!mcpJson.servers)
- mcpJson.servers = {};
-
- mcpJson.servers['playwright-test'] = {
- type: 'stdio',
- command: commonMcpServers.playwrightTest.command,
- args: commonMcpServers.playwrightTest.args,
- cwd: '${workspaceFolder}',
- };
- await writeFile(mcpJsonPath, JSON.stringify(mcpJson, null, 2), 'πŸ”§', 'mcp configuration');
-
- initRepoDone();
-}
-
-export async function initOpencodeRepo(config: FullConfigInternal, projectName: string) {
- await initRepo(config, projectName);
-
- const agents = await loadAgents();
-
- await fs.promises.mkdir('.opencode/prompts', { recursive: true });
- for (const agent of agents) {
- const prompt = [agent.instructions];
- prompt.push('');
- prompt.push(...agent.examples.map(example => `<example>${example}</example>`));
- await writeFile(`.opencode/prompts/playwright-test-${agent.header.name}.md`, prompt.join('\n'), 'πŸ€–', 'agent definition');
- }
- await writeFile('opencode.json', saveAsOpencodeJson(agents), 'πŸ”§', 'opencode configuration');
-
- initRepoDone();
-}
-
const coveragePrompt = (seedFile: string) => `
# Produce test coverage
@@ -356,7 +393,7 @@ Parameters:
- Seed file (optional): the seed file to use, defaults to ${path.relative(process.cwd(), seedFile)}
- Test plan file (optional): the test plan file to write, under specs/ folder.
-1. Call #planner subagent with prompt:
+1. Call #playwright-test-planner subagent with prompt:
<plan>
<task><!-- the task --></task>
@@ -364,7 +401,7 @@ Parameters:
<plan-file><!-- test plan file --></plan-file>
</plan>
-2. For each test case from the test plan file (1.1, 1.2, ...), Call #generator subagent with prompt:
+2. For each test case from the test plan file (1.1, 1.2, ...), Call #playwright-test-generator subagent with prompt:
<generate>
<test-file><!-- Name of the file to save the test into, should be unique for test --></test-file>
@@ -374,7 +411,7 @@ Parameters:
<body><!-- Test case content including steps and expectations --></body>
</generate>
-3. Call #healer subagent with prompt:
+3. Call #playwright-test-healer subagent with prompt:
<heal>Run all tests and fix the failing ones one after another.</heal>
`;
diff --git a/packages/playwright/src/agents/generator.md b/packages/playwright/src/agents/generator.md
index 3e83d145e7..4a313bf56d 100644
--- a/packages/playwright/src/agents/generator.md
+++ b/packages/playwright/src/agents/generator.md
@@ -1,12 +1,11 @@
---
-name: generator
+name: playwright-test-generator
description: Use this agent when you need to create automated browser tests using Playwright
model: sonnet
color: blue
tools:
- - ls
- - grep
- read
+ - search
- playwright-test/browser_click
- playwright-test/browser_drag
- playwright-test/browser_evaluate
diff --git a/packages/playwright/src/agents/healer.md b/packages/playwright/src/agents/healer.md
index ee07a6cb17..92e08e86fd 100644
--- a/packages/playwright/src/agents/healer.md
+++ b/packages/playwright/src/agents/healer.md
@@ -1,12 +1,11 @@
---
-name: healer
+name: playwright-test-healer
description: Use this agent when you need to debug and fix failing Playwright tests
-color: red
model: sonnet
+color: red
tools:
- - ls
- - grep
- read
+ - search
- write
- edit
- playwright-test/browser_console_messages
diff --git a/packages/playwright/src/agents/planner.md b/packages/playwright/src/agents/planner.md
index 192962b296..ae58a7bfeb 100644
--- a/packages/playwright/src/agents/planner.md
+++ b/packages/playwright/src/agents/planner.md
@@ -1,12 +1,11 @@
---
-name: planner
+name: playwright-test-planner
description: Use this agent when you need to create comprehensive test plan for a web application or website
model: sonnet
color: green
tools:
- - ls
- - grep
- read
+ - search
- write
- playwright-test/browser_click
- playwright-test/browser_close
diff --git a/packages/playwright/src/program.ts b/packages/playwright/src/program.ts
index cd9c8e801b..707550c15a 100644
--- a/packages/playwright/src/program.ts
+++ b/packages/playwright/src/program.ts
@@ -37,7 +37,7 @@ import { ServerBackendFactory, runMainBackend } from './mcp/sdk/exports';
import { TestServerBackend } from './mcp/test/testBackend';
import { decorateCommand } from './mcp/program';
import { setupExitWatchdog } from './mcp/browser/watchdog';
-import { initClaudeCodeRepo, initOpencodeRepo, initVSCodeRepo } from './agents/generateAgents';
+import { ClaudeGenerator, OpencodeGenerator, VSCodeGenerator, AgentGenerator } from './agents/generateAgents';
import type { ConfigCLIOverrides } from './common/ipc';
import type { TraceMode } from '../types/test';
@@ -183,18 +183,20 @@ function addInitAgentsCommand(program: Command) {
const command = program.command('init-agents');
command.description('Initialize repository agents');
const option = command.createOption('--loop <loop>', 'Agentic loop provider');
- option.choices(['vscode', 'claude', 'opencode']);
+ option.choices(['vscode', 'claude', 'opencode', 'generic']);
command.addOption(option);
command.option('-c, --config <file>', `Configuration file to find a project to use for seed test`);
command.option('--project <project>', 'Project to use for seed test');
command.action(async opts => {
const config = await loadConfigFromFile(opts.config);
if (opts.loop === 'opencode') {
- await initOpencodeRepo(config, opts.project);
+ await OpencodeGenerator.init(config, opts.project);
} else if (opts.loop === 'vscode') {
- await initVSCodeRepo(config, opts.project);
+ await VSCodeGenerator.init(config, opts.project);
} else if (opts.loop === 'claude') {
- await initClaudeCodeRepo(config, opts.project);
+ await ClaudeGenerator.init(config, opts.project);
+ } else if (opts.loop === 'generic') {
+ await AgentGenerator.init(config, opts.project);
} else {
command.help();
return;