INNER CODE UNIT · Python

run_with_kitsune

wgpsec/AboutSecurity · scripts/bench-skill.py:136

def run_with_kitsune(prompt: str, timeout: int = 120, **kwargs) -> dict:
    """Run a prompt via kitsune agent (placeholder for custom executor)."""
    # Users can extend this for kitsune or other agents
    raise NotImplementedError("Kitsune executor not yet implemented. Use --executor claude")


def build_with_skill_prompt(eval_prompt: str, skill_body: str, skill_name: str) -> str:
    """Build prompt with skill content injected."""
    return f"""你是一个 AI 安全测试 Agent。你有以下安全测试方法论可以参考:

<skill name="{skill_name}">
{skill_body}
</skill>

请根据以上方法论回答以下问题。你的回答应该体现系统化的方法论,而不是随机尝试。

{eval_prompt}"""

View source record →

📰 Research Paper
Loading…
⏳ Fetching content…