The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to orchestrate multiple system profiling tools, including valgrind, perf, and objdump, which is necessary for its function as a performance monitor.
[COMMAND_EXECUTION]: In the tracemalloc stage, the skill dynamically generates a temporary Python wrapper script. This script utilizes runpy and exec() to execute the target benchmark code within a monitored environment, a standard practice for non-intrusive memory profiling.
[PROMPT_INJECTION]: The skill performs autodiscovery of benchmark tests by scanning the target repository for files matching test_*.py. This functionality constitutes an indirect prompt injection surface as it could lead to the execution of malicious code if the scanned repository is compromised.
Ingestion points: Filesystem scanning of the user-provided --root directory in scripts/perf_benchmark/support.py.
Boundary markers: None; the skill assumes the local repository content is trusted by the user.
Capability inventory: Use of pytest and valgrind to execute discovered test files in scripts/perf_benchmark_pipeline.py.
Sanitization: None; discovered file paths are passed directly to subprocess execution commands.

perf-benchmark