The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the Playwright framework and the manual download of the Chromium browser binary via the command playwright install chromium to perform automation tasks.
[COMMAND_EXECUTION]: The skill operates through several Python scripts (e.g., linkedin_engage.py, universal_scraper.py, stats_tracker.py) that execute shell commands to automate browser sessions, interact with local files, and perform network operations.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8).
Ingestion points: Data is ingested from external websites via universal_scraper.py and from the LinkedIn feed via linkedin_engage.py.
Boundary markers: No explicit delimiters or instructions are used to distinguish between system instructions and untrusted data from the browser.
Capability inventory: The skill can perform sensitive write actions, including publishing LinkedIn posts, applying for jobs, and sending notifications via Telegram through stats_tracker.py.
Sanitization: There is no evidence of sanitization or filtering of the scraped content before it is passed to the OpenAI API (GPT-4o-mini) for generating comments or cover letters.
[DATA_EXPOSURE]: The skill's architecture requires the manual extraction and provision of sensitive session cookies (LINKEDIN_LI_AT, TWITTER_AUTH_TOKEN, TWITTER_CT0). While this is a functional requirement for the skill's purpose, the handling of these high-value credentials by an AI agent increases the risk of accidental exposure or misuse.

ghost-browser