The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands and subprocesses to facilitate benchmarking and measurement tasks. The assets/examples/verify.sh script is used to run validation oracles and benchmarks, and assets/examples/constructs/measurement.py invokes the Python interpreter via subprocess.run to measure module import times using the -X importtime flag. These executions are standard for performance diagnostics and are scoped to the development environment.- [DYNAMIC_EXECUTION]: The diagnostic scripts use dynamic execution and inspection techniques to analyze code efficiency. This includes registering sys.monitoring callbacks to count function starts and instruction execution, as well as using the dis module to examine bytecode specialization. These are legitimate activities for a performance auditing tool.- [INDIRECT_PROMPT_INJECTION]: By processing user-provided Python code for optimization, the skill interacts with untrusted data that could influence its execution path. However, the skill provides built-in safeguards by requiring the use of a 'differential oracle'—a test that compares the baseline and optimized code on various inputs to ensure behavior remains identical, which helps detect and prevent the introduction of malicious or incorrect logic.

optimize-python-code