The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external URLs supplied by the user or found during exploration. An attacker-controlled webpage could attempt to influence the agent's behavior via embedded instructions.
Ingestion points: The webpage2pdf/cli.py and webpage2pdf/core.py scripts use driver.get(url) to load external content.
Boundary markers: None detected; the skill does not explicitly warn the agent to ignore instructions embedded in the target webpage.
Capability inventory: The skill has access to the filesystem (via save_pdf) and network (via Selenium/Chrome).
Sanitization: No explicit sanitization or filtering of the webpage content is performed before rendering, although the final output is a rasterized image-based PDF, which prevents direct text-based injection into subsequent agent steps unless OCR is applied.
[EXTERNAL_DOWNLOADS]: The skill relies on selenium >= 4.6, which includes Selenium Manager. This utility automatically downloads the matching chromedriver executable from official Google servers when the skill is run.
[DYNAMIC_EXECUTION]: In webpage2pdf/core.py, the skill explicitly disables Pillow's protection against decompression bomb attacks by setting Image.MAX_IMAGE_PIXELS = None. While the code comment notes this is intended for handling large screenshots generated by Chrome, it removes a security guardrail that prevents resource exhaustion from malicious or malformed image data.
[COMMAND_EXECUTION]: The skill provides a CLI wrapper (webpage2pdf) and instructions for local installation via pip install -e. It uses the Bash tool to execute Chrome and driver-related commands.

webpage-to-pdf