Chrome Profile Agent

Playwright Chrome Profile Agent

The Chrome Profile Agent connects Google Antigravity to your active Google Chrome profile on macOS using Playwright MCP (@playwright/mcp). It allows AI agents to interact with logged-in web applications (LinkedIn, GitHub, Google Workspaces) without requiring re-authentication.

Agent Instructions & Rules

1. Resume File Size Limit Rule

• Strict Requirement: The tailored PDF resume (Raghav Ganesh Resume(.net)-new.pdf) MUST stay strictly under 2 MB (target size: ~1.5 MB).
• Enforcement: PyMuPDF script tailor_resume_helper.py uses Base-14 standard fonts and stream deflation (deflate=True, garbage=4, clean=True) to purge unreferenced font objects and keep the output file size strictly below 2 MB.

2. Comprehensive Easy Apply Form Filling Rules

When navigating LinkedIn Easy Apply modal dialogs (.jobs-easy-apply-modal), the agent MUST systematically check and fill:

• Resume Upload: Attach Raghav Ganesh Resume(.net)-new.pdf (< 2MB).
• Textboxes & Number Inputs:
  • Years of Experience / Skill ratings: Default to 4 years.
  • Notice Period fields: Fill 15 days or Immediate.
  • Compensation / CTC fields: Fill expected CTC values.
• Dropdowns & Comboboxes:
  • Native <select> elements: Inspect <option> list and choose matching choice ("Yes", "Bachelor's", "Full-time", "15 days").
  • ARIA comboboxes ([role="combobox"]): Click to expand options and select first valid choice.
• Radio Buttons & Checkboxes:
  • Work Authorization: Select Yes.
  • Sponsorship: Select No (or Yes if required).
  • Terms / Agreements: Check all mandatory checkboxes.

Operating Modes

Mode 1: Automated LinkedIn Easy Apply Workflow

Run the end-to-end Easy Apply automation script:

# 1. Test Easy Apply workflow in dry-run mode (stops before final submission)
node scripts/easy_apply_runner.js --search ".NET Full Stack Developer" --dry-run

# 2. Execute full Easy Apply workflow with auto-submission & Google Sheets logging
node scripts/easy_apply_runner.js --search ".NET Developer"

Workflow Steps executed by easy_apply_runner.js:

1. Navigates to LinkedIn Jobs with f_AL=true (Easy Apply filter).
2. Selects an Easy Apply job listing and extracts full Job Description (jdText).
3. Executes tailor_resume_helper.py to update Raghav Ganesh Resume(.net)-new.pdf (strictly < 2MB).
4. Clicks .jobs-apply-button to open the Easy Apply modal.
5. Uploads tailored resume PDF, fills textboxes, handles <select> dropdowns and ARIA comboboxes, and answers radio buttons.
6. Clicks Next $\rightarrow$ Review $\rightarrow$ Submit application.
7. Logs the submission directly to Job Tracker - Raghav (Table1, range F:M).

Mode 2: CDP Connection (Active Chrome Session)

Attaches directly to your running Google Chrome browser over Chrome Developer Protocol (CDP):

1. Check CDP Port Status:

   curl -s http://localhost:9222/json/version
   

2. Launch Chrome with CDP:

   ./scripts/launch_chrome_cdp.sh
   

3. Run Playwright MCP:

   npx -y @playwright/mcp --cdp-endpoint http://localhost:9222
   

Mode 3: Cloned Profile Session (Co-existence with Open Chrome)

If Chrome is open and cannot be restarted in CDP mode, Playwright syncs session state (Cookies, Web Data, Local Storage, Preferences, Login Data) to ~/.chrome-playwright-profile to bypass Chrome's SingletonLock:

node scripts/chrome_runner.js --search ".NET Full Stack Developer"

Workflow Guide for AI Agents

When assisting a user with job application tasks:

1. Ensure the tailored PDF size remains strictly < 2 MB.
2. Execute scripts/easy_apply_runner.js with --search for the requested role.
3. Automatically handle textboxes, select dropdowns, comboboxes, radio buttons, and resume upload.
4. All completed applications will automatically log to Job Tracker - Raghav in Google Sheets (Table1 range F:M).