The Agent Skills Directory

[COMMAND_EXECUTION]: Provides implementations for establishing system persistence by modifying Linux crontab files (rpa-scheduling.md) and creating Windows Scheduled Tasks via schtasks.exe (rpa-scheduling.md).
[COMMAND_EXECUTION]: Includes patterns for full desktop automation, including mouse manipulation, keyboard simulation, and window control using the pyautogui and pywinauto libraries (rpa-desktop.md).
[COMMAND_EXECUTION]: Employs the subprocess and psutil modules to programmatically launch, monitor, and terminate system processes (rpa-desktop.md).
[COMMAND_EXECUTION]: Leverages Playwright's evaluate and add_init_script methods to execute custom JavaScript within the browser context for scraping and stealth purposes (rpa-browser-automation.md, rpa-data-extraction.md).
[DATA_EXFILTRATION]: Includes capabilities for sending extracted data to external endpoints through SMTP email (rpa-documents.md), Slack webhooks, and Telegram bot APIs (rpa-error-handling.md, rpa-observability.md).
[DATA_EXFILTRATION]: Integrates with the 2captcha.com third-party service, which involves transmitting image and site metadata to external servers for automated CAPTCHA solving (rpa-form-automation.md).
[EXTERNAL_DOWNLOADS]: Documentation instructs on fetching the uv package manager installation script from astral.sh using a curl | sh pattern (rpa-cicd.md).
[COMMAND_EXECUTION]: Presents an indirect prompt injection attack surface.
Ingestion points: Processes untrusted data from web pages via Playwright (rpa-data-extraction.md) and emails via IMAP (rpa-documents.md).
Boundary markers: Examples lack delimiters or specific instructions to ignore malicious commands embedded in processed content.
Capability inventory: Possesses powerful tools for desktop control (pyautogui), persistence (crontab), and network transmission (aiosmtplib).
Sanitization: No explicit sanitization or validation of ingested content before it is processed or used in further actions.

rpa