init agent project

common/prompt_guard.py

@@ -0,0 +1,66 @@
+"""
+防提示词注入模块
+
+在模型调用前检测并拦截潜在的提示词注入攻击。
+"""
+
+import re
+import logging
+from typing import Optional
+
+logger = logging.getLogger("adk.agent")
+
+# ============================================================
+# 常见提示词注入模式
+# ============================================================
+_INJECTION_PATTERNS = [
+    # 角色扮演/身份覆盖
+    r"ignore\s+(all\s+)?previous\s+(instructions?|prompts?|rules?)",
+    r"forget\s+(all\s+)?previous\s+(instructions?|prompts?|rules?)",
+    r"disregard\s+(all\s+)?previous\s+(instructions?|prompts?|rules?)",
+    r"you\s+are\s+now\s+a",
+    r"pretend\s+(you\s+are|to\s+be)",
+    r"act\s+as\s+(if\s+you\s+(are|were)|a|an)",
+    r"roleplay\s+as",
+    r"你是一个",
+    r"假装你是",
+    r"扮演一个",
+    # 指令泄露
+    r"(show|reveal|display|print|output|dump)\s+(me\s+)?(your|the)?\s*system\s*(prompt|instructions?|rules?|config)",
+    r"(show|reveal|display|print|output|dump)\s+(me\s+)?your\s+(prompt|instructions?|rules?|config)",
+    r"what\s+(are|is)\s+your\s+(instructions?|prompts?|rules?|system\s*prompt)",
+    r"repeat\s+(your|the|back)\s+(instructions?|prompts?|system)",
+    r"(显示|输出|打印|告诉我|泄露)\s*(你的|系统|隐藏)",
+    # 分隔符注入
+    r"---\s*(system|instruction|prompt)",
+    r"###\s*(system|instruction|prompt)",
+    r"<\|im_start\|>",
+    r"<\|im_end\|>",
+    # 越狱尝试
+    r"(jailbreak|dan\s+mode|developer\s+mode)\b",
+    r"bypass\s+(safety|filter|security|restriction)",
+    r"no\s+(safety|ethical|moral|content)\s+(filter|restriction|limit)",
+]
+
+_COMPILED_PATTERNS = [re.compile(p, re.IGNORECASE) for p in _INJECTION_PATTERNS]
+
+
+def check_prompt_injection(text: str) -> Optional[str]:
+    """
+    检测文本中是否包含提示词注入。
+
+    Args:
+        text: 待检测的用户输入文本
+
+    Returns:
+        如果检测到注入，返回匹配到的模式描述；否则返回 None
+    """
+    if not text:
+        return None
+
+    for pattern in _COMPILED_PATTERNS:
+        match = pattern.search(text)
+        if match:
+            return match.group(0)
+
+    return None