The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the 'graphifyy' package from PyPI. This name is a slight variation of the skill's name 'graphify', which is a common signal for typosquatting or supply-chain risk.
[COMMAND_EXECUTION]: The skill uses shell commands to clone external repositories, install software, and execute system-level operations based on user input. It also establishes persistence by installing a git post-commit hook.
[REMOTE_CODE_EXECUTION]: The instructions involve generating Python scripts dynamically and executing them via the command line. It also saves the interpreter state to a local file for subsequent automated use of the installed package.
[PROMPT_INJECTION]: The skill exhibits a high attack surface for indirect prompt injection by reading content from local folders, GitHub repositories, YouTube URLs, Twitter posts, and PDF documents. It processes this potentially untrusted data using 'general-purpose' subagents which possess Write and Bash capabilities, and it does not implement strong boundary markers or explicit sanitization steps.

graphify