The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/zurl_api.py executes a hardcoded system command to set the terminal character encoding on Windows platforms.\n
Evidence: subprocess.run("chcp 65001", shell=True, capture_output=True) in scripts/zurl_api.py.\n
Context: This is a static environment initialization step and is not user-controllable.\n- [PROMPT_INJECTION]: The skill processes data from external websites, creating a surface for indirect prompt injection.\n
Ingestion points: Page titles and descriptions are fetched from remote URLs via the metadata command in scripts/zurl_api.py.\n
Boundary markers: The script returns all API data to the agent as structured JSON, providing a clear boundary between external data and agent instructions.\n
Capability inventory: The skill is capable of making network requests to a user-defined API endpoint and reading local configuration files from the home directory.\n
Sanitization: External content is sanitized through standard JSON serialization before being processed by the agent.\n- [EXTERNAL_DOWNLOADS]: The skill instructions require the installation of an external dependency for environment management.\n
Evidence: pip install python-dotenv listed in SKILL.md.\n
Context: This is a standard and well-known library used for secure configuration management.

zurl