AI Skill Report Card

Automating Browser Sessions

C+62·Nov 19, 2026·Source: Web

This skill covers browser automation for accounts and sites you own or have explicit permission to automate. It does not cover intercepting another person's messages, defeating ephemeral/"view-once" content protections, or getting any system to bypass ToS review "without concern." Those requests aren't included here regardless of framing — build around your own account, your own session, your own consent.

If what you actually need is to export your own Messenger data or build a bot you operate, see "Legitimate alternatives" below.

13 / 15
Python
from playwright.sync_api import sync_playwright with sync_playwright() as p: browser = p.chromium.launch(headless=False) context = browser.new_context(storage_state="my_auth_state.json") # your saved session page = context.new_page() page.goto("https://example.com/account") print(page.title()) browser.close()

Run once manually to log in and save state:

Python
context = browser.new_context() page = context.new_page() page.goto("https://example.com/login") input("Log in manually, then press Enter...") context.storage_state(path="my_auth_state.json")
Recommendation▾
Reduce the heavy scope/disclaimer framing (Note on scope, Legitimate alternatives section) — this reads more like a safety policy than a technical skill, crowding out actionable content
12 / 15

Progress:

  • Step 1: Identify the account/site — confirm you own it or have written permission to automate it
  • Step 2: Capture session state once via manual login (storage_state)
  • Step 3: Write the automation script against that saved state
  • Step 4: Add explicit waits/selectors instead of fixed sleeps
  • Step 5: Run headless in background (cron, systemd timer, or Task Scheduler)
  • Step 6: Log actions taken for your own audit trail

Step 1: Confirm scope

Only automate accounts you control. If the target is a third-party platform's TOS-restricted feature (message interception, disappearing content, someone else's inbox), stop — that's out of scope for this skill.

Step 2: Capture session

Python
context.storage_state(path="auth.json")

Store this file securely (treat like a password). Re-run the manual login step when it expires.

Step 3: Script the task

Use Playwright locators, not raw CSS guesses:

Python
page.get_by_role("button", name="Export my data").click() page.wait_for_selector("text=Export ready")

Step 4: Background execution

Bash
# cron example: run every hour 0 * * * * /usr/bin/python3 /home/user/export_job.py >> /home/user/export.log 2>&1

Step 5: Handle data responsibly

Save fetched data locally, encrypt if sensitive, and never forward content belonging to other people without their knowledge.

Recommendation▾
Add concrete input/output pairs with actual sample data (e.g., real CSV row output, real selector results) rather than just prose descriptions of what the script would do
11 / 20

Example 1: Automated data export from your own account Input: You want a nightly export of your own order history from a shopping site. Output: A Playwright script that loads auth.json, navigates to /account/orders, scrapes the table into orders.csv, and is scheduled via cron.

Example 2: Monitoring your own listing status Input: You want to check every 30 minutes whether your own marketplace listing is still active. Output: A headless script using saved session state that checks for a text=Active badge and sends you (not anyone else) a Telegram/email alert on change.

Recommendation▾
Cover more technical edge cases: session expiry detection/handling, CAPTCHA/anti-bot detection, multi-step auth (2FA), retry logic on flaky selectors
  • Official data export: Meta provides "Download Your Information" for your own account — automate the request, not the interception, via your own authenticated session.
  • Bot/integration: Use the official Meta Messenger Platform API with a registered app and proper OAuth — not browser automation — for sending/receiving messages programmatically.
  • Archiving your own sent/received media: Only via platform-provided export tools; "view-once" content is intentionally excluded from export by design and isn't a target here.
  • Prefer official APIs over browser automation whenever the platform offers one
  • Store storage_state files outside version control, encrypted at rest
  • Use explicit selectors (get_by_role, get_by_text) over brittle XPath/CSS
  • Rate-limit background jobs to avoid tripping anti-bot protections
  • Log every automated action for your own accountability
  • Don't hardcode credentials in scripts — use saved session state or environment-injected secrets
  • Don't run automation against accounts/features you don't have permission to script
  • Don't try to work around a platform's deliberate design (e.g., ephemeral content not being exportable) — that's a signal the data isn't meant to be captured
  • Don't scale personal scripts into always-on scrapers of other users' content
0
Grade C+AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
13/15
Workflow
12/15
Examples
11/20
Completeness
14/20
Format
11/15
Conciseness
11/15