The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/compare_typecheckers.py executes external type checkers using subprocess.run. The implementation correctly avoids using shell=True and passes arguments as a list, which is a secure method for executing external processes that prevents command injection.
[EXTERNAL_DOWNLOADS]: The skill documentation and code indicate the use of uvx to fetch and run type checkers from official package registries on demand. These tools originate from well-known sources (such as Meta and Astral) and represent standard developer workflow automation.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted Python code provided by the user for analysis. However, the risk is negligible because the tools perform static analysis without executing the input code, and the script does not interpolate the untrusted content into sensitive system prompts.
Ingestion points: User-provided Python code via the file argument or the -c/--code flag in scripts/compare_typecheckers.py.
Boundary markers: None explicitly defined in the prompts to isolate processed data.
Capability inventory: The skill has the capability to execute subprocesses and write results to local files specified by the user.
Sanitization: The input code is not sanitized as it is intended for static analysis by third-party type checkers.

compare-python-typecheckers