The Agent Skills Directory

[SAFE]: The skill implements a structured workflow for instrumenting Python agents with Monte Carlo Observability. It adheres to the principle of least privilege by requiring explicit user confirmation before any file modification (dependency files, source code, or environment files).
[SAFE]: Sensitive credentials (MCD API tokens) are handled securely. The skill documentation and implementation (e.g., in references/setup-template.md and references/troubleshooting.md) explicitly forbid reading, logging, or echoing the values of these tokens, mandating only presence-only checks.
[SAFE]: Remote information fetching is limited to PyPI (pypi.org), which is a well-known and trusted service. The fetch_sdk_docs.py script retrieves metadata and README content to identify supported instrumentors and version pins without executing any retrieved content.
[SAFE]: Helper scripts (detect_libraries.py, fetch_sdk_docs.py) perform discovery and metadata retrieval in a transparent manner. The use of ast.parse for existing-setup detection is a safe alternative to dynamic execution.
[SAFE]: Subprocess calls in the tests/ directory are standard for testing CLI-like scripts within a development environment and are limited to executing the skill's own scripts using the local Python interpreter.

monte-carlo-instrument-agent