The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill manages a local Python virtual environment and installs dependencies from the standard Python Package Index (PyPI). It uses a lock file with cryptographic hashes (requirements.lock) to ensure the integrity of the installed packages.
[COMMAND_EXECUTION]: Local shell commands are used via the subprocess module in run.py and setup_environment.py for managing the execution environment and running internal scripts. These operations are restricted to the skill's directory and do not take unvalidated user input.
[INDIRECT_PROMPT_INJECTION]: The nlp_analyze.py script can fetch content from external URLs for natural language processing analysis. While this creates a surface where untrusted data enters the agent's context, the script includes SSRF protections by validating that hostnames do not resolve to private, reserved, or loopback IP addresses.
[DYNAMIC_EXECUTION]: The runner script (run.py) dynamically selects which internal Python script to execute based on user commands. It implements security checks to ensure the requested script path does not escape the permitted scripts directory, preventing directory traversal attacks.

blog-google