The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/http_benchmark.py executes external load-testing tools (wrk or ab) using the subprocess.run module.
Evidence: Found in functions run_wrk and run_ab where command lists are constructed and executed.
Mitigation: The script uses argument lists instead of raw shell strings and avoids the shell=True parameter, which prevents standard shell command injection vulnerabilities. It also validates the target URL scheme and hostname before execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from external benchmarking tools, which are influenced by responses from the target URL.
Ingestion points: The skill captures stdout and stderr from the execution of external binaries in scripts/http_benchmark.py.
Boundary markers: No explicit boundary markers or delimiters are used to wrap the external tool output before processing.
Capability inventory: The skill utilizes subprocess.run to call external binaries and performs network-based load testing against user-provided URLs.
Sanitization: The script uses strict regular expression patterns to extract specific numerical metrics (RPS, latency percentiles) from the tool output, which minimizes the risk of arbitrary text in the tool output being misinterpreted by the agent.

http-load-profiler