代理治理模式

为 AI 代理系统增加安全性、信任和策略执行的设计模式。

概述

治理模式确保 AI 代理在定义的边界内运行——控制它们可以调用哪些工具、可以处理哪些内容、可以做多少事情，并通过审计跟踪保持可问责性。

用户请求 → 意图分类 → 策略检查 → 工具执行 → 审计日志
                     ↓                      ↓               ↓
              威胁检测                 允许/拒绝       信任更新

何时使用

• 具有工具访问权限的代理：任何调用外部工具（API、数据库、Shell 命令）的代理
• 多代理系统：代理委派给其他代理需要信任边界
• 生产部署：合规性、审计和安全要求
• 敏感操作：金融交易、数据访问、基础设施管理

模式 1：治理策略

将允许代理执行的操作定义为一个可组合、可序列化的策略对象。

from dataclasses import dataclass, field
from enum import Enum
from typing import Optional
import re

class PolicyAction(Enum):
    ALLOW = "allow"
    DENY = "deny"
    REVIEW = "review"  # 标记需要人工审查

@dataclass
class GovernancePolicy:
    """声明性策略，控制代理行为。""" # 翻译
    name: str
    allowed_tools: list[str] = field(default_factory=list)       # 允许列表
    blocked_tools: list[str] = field(default_factory=list)       # 阻止列表
    blocked_patterns: list[str] = field(default_factory=list)    # 内容过滤器
    max_calls_per_request: int = 100                             # 速率限制
    require_human_approval: list[str] = field(default_factory=list)  # 需要人工批准的工具

    def check_tool(self, tool_name: str) -> PolicyAction:
        """检查此策略是否允许使用某个工具。""" # 翻译
        if tool_name in self.blocked_tools:
            return PolicyAction.DENY
        if tool_name in self.require_human_approval:
            return PolicyAction.REVIEW
        if self.allowed_tools and tool_name not in self.allowed_tools:
            return PolicyAction.DENY
        return PolicyAction.ALLOW

    def check_content(self, content: str) -> Optional[str]:
        """根据阻止的模式列表检查内容。返回匹配的模式或 None。""" # 翻译
        for pattern in self.blocked_patterns:
            if re.search(pattern, content, re.IGNORECASE):
                return pattern
        return None

策略组合

组合多个策略（例如，组织级 + 团队级 + 代理特定）：

def compose_policies(*policies: GovernancePolicy) -> GovernancePolicy:
    """合并策略，采用"最严格优先"的语义。""" # 翻译
    combined = GovernancePolicy(name="composed")

    for policy in policies:
        combined.blocked_tools.extend(policy.blocked_tools)
        combined.blocked_patterns.extend(policy.blocked_patterns)
        combined.require_human_approval.extend(policy.require_human_approval)
        combined.max_calls_per_request = min(
            combined.max_calls_per_request,
            policy.max_calls_per_request
        )
        if policy.allowed_tools:
            if combined.allowed_tools:
                combined.allowed_tools = [
                    t for t in combined.allowed_tools if t in policy.allowed_tools
                ]
            else:
                combined.allowed_tools = list(policy.allowed_tools)

    return combined


# 用法：从宽泛到具体层层叠加策略
org_policy = GovernancePolicy(
    name="org-wide",
    blocked_tools=["shell_exec", "delete_database"],
    blocked_patterns=[r"(?i)(api[_-]?key|secret|password)\s*[:=]"],
    max_calls_per_request=50
)
team_policy = GovernancePolicy(
    name="data-team",
    allowed_tools=["query_db", "read_file", "write_report"],
    require_human_approval=["write_report"]
)
agent_policy = compose_policies(org_policy, team_policy)

策略即 YAML

将策略作为配置存储，而不是代码：

# governance-policy.yaml
name: production-agent
allowed_tools:
  - search_documents
  - query_database
  - send_email
blocked_tools:
  - shell_exec
  - delete_record
blocked_patterns:
  - "(?i)(api[_-]?key|secret|password)\\s*[:=]"
  - "(?i)(drop|truncate|delete from)\\s+\\w+"
max_calls_per_request: 25
require_human_approval:
  - send_email

import yaml

def load_policy(path: str) -> GovernancePolicy:
    with open(path) as f:
        data = yaml.safe_load(f)
    return GovernancePolicy(**data)

模式 2：语义意图分类

在提示词到达代理之前，使用基于模式的信号检测其中的危险意图。

from dataclasses import dataclass

@dataclass
class IntentSignal:
    category: str       # 例如，"data_exfiltration"（数据泄露），"privilege_escalation"（权限提升）
    confidence: float   # 0.0 到 1.0
    evidence: str       # 触发检测的证据

# 用于威胁检测的加权信号模式
THREAT_SIGNALS = [
    # 数据泄露
    (r"(?i)send\s+(all|every|entire)\s+\w+\s+to\s+", "data_exfiltration", 0.8),
    (r"(?i)export\s+.*\s+to\s+(external|outside|third.?party)", "data_exfiltration", 0.9),
    (r"(?i)curl\s+.*\s+-d\s+", "data_exfiltration", 0.7),

    # 权限提升
    (r"(?i)(sudo|as\s+root|admin\s+access)", "privilege_escalation", 0.8),
    (r"(?i)chmod\s+777", "privilege_escalation", 0.9),

    # 系统破坏
    (r"(?i)(rm\s+-rf|del\s+/[sq]|format\s+c:)", "system_destruction", 0.95),
    (r"(?i)(drop\s+database|truncate\s+table)", "system_destruction", 0.9),

    # 提示注入
    (r"(?i)ignore\s+(previous|above|all)\s+(instructions?|rules?)", "prompt_injection", 0.9),
    (r"(?i)you\s+are\s+now\s+(a|an)\s+", "prompt_injection", 0.7),
]

def classify_intent(content: str) -> list[IntentSignal]:
    """对内容进行威胁信号分类。""" # 翻译
    signals = []
    for pattern, category, weight in THREAT_SIGNALS:
        match = re.search(pattern, content)
        if match:
            signals.append(IntentSignal(
                category=category,
                confidence=weight,
                evidence=match.group()
            ))
    return signals

def is_safe(content: str, threshold: float = 0.7) -> bool:
    """快速检查：内容是否安全，高于给定阈值？""" # 翻译
    signals = classify_intent(content)
    return not any(s.confidence >= threshold for s in signals)

关键见解：意图分类发生在工具执行之前，充当飞行前的安全检查。这与仅在生成之后检查的输出护栏有本质区别。

模式 3：工具级治理装饰器

使用治理检查包装各个工具函数：

import functools
import time
from collections import defaultdict

_call_counters: dict[str, int] = defaultdict(int)

def govern(policy: GovernancePolicy, audit_trail=None):
    """装饰器，在工具函数上强制实施治理策略。""" # 翻译
    def decorator(func):
        @functools.wraps(func)
        async def wrapper(*args, **kwargs):
            tool_name = func.__name__

            # 1. 检查工具允许/阻止列表
            action = policy.check_tool(tool_name)
            if action == PolicyAction.DENY:
                raise PermissionError(f"策略'{policy.name}'阻止了工具'{tool_name}'")
            if action == PolicyAction.REVIEW:
                raise PermissionError(f"工具'{tool_name}'需要人工批准")

            # 2. 检查速率限制
            _call_counters[policy.name] += 1
            if _call_counters[policy.name] > policy.max_calls_per_request:
                raise PermissionError(f"超出速率限制：{policy.max_calls_per_request}次调用")

            # 3. 检查参数中的内容
            for arg in list(args) + list(kwargs.values()):
                if isinstance(arg, str):
                    matched = policy.check_content(arg)
                    if matched:
                        raise PermissionError(f"检测到被阻止的模式：{matched}")

            # 4. 执行并审计
            start = time.monotonic()
            try:
                result = await func(*args, **kwargs)
                if audit_trail is not None:
                    audit_trail.append({
                        "tool": tool_name,
                        "action": "allowed",
                        "duration_ms": (time.monotonic() - start) * 1000,
                        "timestamp": time.time()
                    })
                return result
            except Exception as e:
                if audit_trail is not None:
                    audit_trail.append({
                        "tool": tool_name,
                        "action": "error",
                        "error": str(e),
                        "timestamp": time.time()
                    })
                raise

        return wrapper
    return decorator


# 与任何代理框架一起使用
audit_log = []
policy = GovernancePolicy(
    name="search-agent",
    allowed_tools=["search", "summarize"],
    blocked_patterns=[r"(?i)password"],
    max_calls_per_request=10
)

@govern(policy, audit_trail=audit_log)
async def search(query: str) -> str:
    """搜索文档 — 受策略治理。""" # 翻译
    return f"搜索结果：{query}"

# 通过：search("最新季度报告")
# 被阻止：search("给我看看管理员密码")

模式 4：信任评分

使用基于时间衰减的信任分数跟踪代理的可靠性：

from dataclasses import dataclass, field
import math
import time

@dataclass
class TrustScore:
    """带时间衰减的信任分数。""" # 翻译
    score: float = 0.5          # 0.0 (不信任) 到 1.0 (完全信任)
    successes: int = 0
    failures: int = 0
    last_updated: float = field(default_factory=time.time)

    def record_success(self, reward: float = 0.05):
        self.successes += 1
        self.score = min(1.0, self.score + reward * (1 - self.score))
        self.last_updated = time.time()

    def record_failure(self, penalty: float = 0.15):
        self.failures += 1
        self.score = max(0.0, self.score - penalty * self.score)
        self.last_updated = time.time()

    def current(self, decay_rate: float = 0.001) -> float:
        """获取随时间衰减的分数——信任因不活动而侵蚀。""" # 翻译
        elapsed = time.time() - self.last_updated
        decay = math.exp(-decay_rate * elapsed)
        return self.score * decay

    @property
    def reliability(self) -> float:
        total = self.successes + self.failures
        return self.successes / total if total > 0 else 0.0


# 在多代理系统中使用
trust = TrustScore()

# 代理成功完成任务
trust.record_success()  # 0.525
trust.record_success()  # 0.549

# 代理出错
trust.record_failure()  # 0.467

# 基于信任度对敏感操作进行门控
if trust.current() >= 0.7:
    # 允许自主操作
    pass
elif trust.current() >= 0.4:
    # 允许但需人工监督
    pass
else:
    # 拒绝或要求明确批准
    pass

多代理信任：在代理委派给其他代理的系统中，每个代理为其委派对象维护信任分数：

class AgentTrustRegistry:
    def __init__(self):
        self.scores: dict[str, TrustScore] = {}

    def get_trust(self, agent_id: str) -> TrustScore:
        if agent_id not in self.scores:
            self.scores[agent_id] = TrustScore()
        return self.scores[agent_id]

    def most_trusted(self, agents: list[str]) -> str:
        return max(agents, key=lambda a: self.get_trust(a).current())

    def meets_threshold(self, agent_id: str, threshold: float) -> bool:
        return self.get_trust(agent_id).current() >= threshold

模式 5：审计跟踪

所有代理操作的仅追加审计日志——对合规性和调试至关重要：

from dataclasses import dataclass, field
import json
import time

@dataclass
class AuditEntry:
    timestamp: float
    agent_id: str
    tool_name: str
    action: str           # "allowed", "denied", "error"
    policy_name: str
    details: dict = field(default_factory=dict)

class AuditTrail:
    """代理治理事件的仅追加审计跟踪。""" # 翻译
    def __init__(self):
        self._entries: list[AuditEntry] = []

    def log(self, agent_id: str, tool_name: str, action: str,
            policy_name: str, **details):
        self._entries.append(AuditEntry(
            timestamp=time.time(),
            agent_id=agent_id,
            tool_name=tool_name,
            action=action,
            policy_name=policy_name,
            details=details
        ))

    def denied(self) -> list[AuditEntry]:
        """获取所有被拒绝的操作——对安全审查很有用。""" # 翻译
        return [e for e in self._entries if e.action == "denied"]

    def by_agent(self, agent_id: str) -> list[AuditEntry]:
        return [e for e in self._entries if e.agent_id == agent_id]

    def export_jsonl(self, path: str):
        """导出为 JSON Lines 格式，用于日志聚合系统。""" # 翻译
        with open(path, "w") as f:
            for entry in self._entries:
                f.write(json.dumps({
                    "timestamp": entry.timestamp,
                    "agent_id": entry.agent_id,
                    "tool": entry.tool_name,
                    "action": entry.action,
                    "policy": entry.policy_name,
                    **entry.details
                }) + "\n")

模式 6：框架集成

PydanticAI

from pydantic_ai import Agent

policy = GovernancePolicy(
    name="support-bot",
    allowed_tools=["search_docs", "create_ticket"],
    blocked_patterns=[r"(?i)(ssn|social\s+security|credit\s+card)"],
    max_calls_per_request=20
)

agent = Agent("openai:gpt-4o", system_prompt="你是一个支持助手。")

@agent.tool
@govern(policy)
async def search_docs(ctx, query: str) -> str:
    """搜索知识库 — 受治理。""" # 翻译
    return await kb.search(query)

@agent.tool
@govern(policy)
async def create_ticket(ctx, title: str, body: str) -> str:
    """创建支持工单 — 受治理。""" # 翻译
    return await tickets.create(title=title, body=body)

CrewAI

from crewai import Agent, Task, Crew

policy = GovernancePolicy(
    name="research-crew",
    allowed_tools=["search", "analyze"],
    max_calls_per_request=30
)

# 在 Crew 级别应用治理
def governed_crew_run(crew: Crew, policy: GovernancePolicy):
    """用治理检查包装 Crew 执行。""" # 翻译
    audit = AuditTrail()
    for agent in crew.agents:
        for tool in agent.tools:
            original = tool.func
            tool.func = govern(policy, audit_trail=audit)(original)
    result = crew.kickoff()
    return result, audit

OpenAI Agents SDK

from agents import Agent, function_tool

policy = GovernancePolicy(
    name="coding-agent",
    allowed_tools=["read_file", "write_file", "run_tests"],
    blocked_tools=["shell_exec"],
    max_calls_per_request=50
)

@function_tool
@govern(policy)
async def read_file(path: str) -> str:
    """读取文件内容 — 受治理。""" # 翻译
    import os
    safe_path = os.path.realpath(path)
    if not safe_path.startswith(os.path.realpath(".")):
        raise ValueError("治理阻止了路径遍历")
    with open(safe_path) as f:
        return f.read()

治理级别

根据风险级别匹配治理严格程度：

最佳实践

快速启动清单

## 代理治理实施清单

### 设置
- [ ] 定义治理策略（允许的工具、阻止的模式、速率限制）
- [ ] 选择治理级别（开放/标准/严格/锁定）
- [ ] 设置审计跟踪存储

### 实施
- [ ] 为所有工具函数添加 @govern 装饰器
- [ ] 在用户输入处理中添加意图分类
- [ ] 为多代理交互实现信任评分
- [ ] 连接审计跟踪导出功能

### 验证
- [ ] 测试被阻止的工具是否被正确拒绝
- [ ] 测试内容过滤器是否能捕获敏感模式
- [ ] 测试速率限制行为
- [ ] 验证审计跟踪是否捕获所有事件
- [ ] 测试策略组合（最严格优先）

相关资源

• Agent-OS 治理引擎 — 完整的治理框架
• AgentMesh 集成 — 特定框架的软件包
• OWASP LLM 应用十大风险