The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/humane_setup.py script utilizes subprocess.run(shell=True) to execute environment remediation commands.
[COMMAND_EXECUTION]: A command injection vulnerability exists in the handling of the token_base configuration setting. Values provided in a project's humane.json file are interpolated into shell commands (e.g., tokens setup-edit {base}) and executed without sanitization, enabling arbitrary code execution via malicious project files.
[DYNAMIC_EXECUTION]: The script performs dynamic module loading by modifying sys.path to import illustrate.py from a relative directory (../../brand-illustrate/scripts), which can be exploited for local code execution.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of third-party components, including the agent-browser package from npm and plugins from the jakubkrehel/skills GitHub repository.
[REMOTE_CODE_EXECUTION]: The inclusion of shell-executed commands derived from untrusted project-local configuration files allows for arbitrary code execution when interacting with a malicious repository.

setup