The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to manage the server lifecycle and run profiling tools.
Evidence: sglang serve, python3 -m sglang.test.run_eval, and python3 -m sglang.test.send_one in SKILL.md.
[COMMAND_EXECUTION]: Employs the pkill command to terminate background server processes.
Evidence: pkill -9 -f "sglang.launch_server\|sglang serve\|sglang.srt" in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface where untrusted user input could influence shell command execution.
Ingestion points: User-supplied values for <gpu_id>, <model>, and <port> in SKILL.md.
Boundary markers: Absent; placeholders are directly interpolated into shell commands without delimiters.
Capability inventory: Execution of shell commands and process management via sglang CLI and Python modules.
Sanitization: None; the instructions do not specify validation or escaping for the user-provided strings.

generate-profile