The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to clone and execute code from an unverified third-party repository (eze-is/web-access). This source is not associated with a verified or trusted organization.
[REMOTE_CODE_EXECUTION]: By recommending the installation and execution of scripts (specifically scripts/check-deps.mjs) from an external GitHub account that is not part of a trusted organization, the skill introduces a risk of remote code execution.
[COMMAND_EXECUTION]: The skill performs extensive command execution using PowerShell, Node.js, and Python. It hardcodes a specific absolute path to a Python executable (C:\Users\louis\mysoft\env\pyenv\codex\Scripts\python.exe) and executes scripts with arguments derived from external file contents.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via the processing of the dois_row.txt file.
Ingestion points: The skill reads dois_row.txt to extract DOIs and subfolder names.
Boundary markers: There are no delimiters or instructions to treat the file content as untrusted data.
Capability inventory: The skill has the capability to execute shell commands (node, python) and perform file system operations (New-Item, Copy-Item).
Sanitization: While the script splits lines by whitespace, the resulting tokens are used directly as command-line arguments and file system paths without validation, potentially allowing for directory traversal or argument injection.

p-pdf-nature-downloader