The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill is designed to process untrusted HTML data which can contain embedded instructions targeting the AI agent.
Ingestion points: The html_to_long_image.py script accepts arbitrary HTML file paths as command-line arguments.
Boundary markers: Absent. The script lacks any delimiters or safety instructions to prevent the agent from interpreting the content of the HTML files.
Capability inventory: The AI agent utilizing this skill typically possesses file system access and command execution capabilities (evidenced by the skill's own use of subprocess).
Sanitization: Absent. The HTML is rendered directly via Playwright without any sanitization or filtering of script tags or malicious metadata.
Unverifiable Dependencies (MEDIUM): The script html_to_long_image.py performs runtime installation of Python packages (playwright, Pillow) using subprocess.check_call([sys.executable, '-m', 'pip', 'install', ...]). This bypasses secure dependency management and exposes the system to supply chain attacks or environment poisoning at runtime.
Command Execution (LOW): The skill documentation and scripts rely on several shell commands (pip, playwright install, chmod +x, open) for installation and operation, increasing the execution footprint.
Missing Components (LOW): The SKILL.md file references html_to_pdf_final.py and html_to_pdf_converter.py, which are not included in the provided file list. This prevents a complete security audit of the skill's full capabilities and creates potential for the agent to attempt to execute non-existent or unverified code.

html-to-pdf