The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the subprocess module to execute local Python scripts, the claude CLI tool, and system utilities such as lsof and kill to manage the benchmarking lifecycle and the local evaluation viewer server.- [PROMPT_INJECTION]: The skill implements an evaluation workflow where user-provided test prompts are passed directly to subagents for execution. This ingestion of untrusted data without explicit sanitization or boundary markers presents a surface for indirect prompt injection, as malicious test cases could attempt to subvert subagent logic or safety constraints.

skill-creator