什么是 自主执行器？

自主执行器是一个先进的自我修复引擎，旨在为 Openclaw Skills 提供完全的运行自主权。它能智能监控任务执行，识别网络、身份验证和依赖关系等各类故障。通过应用指数退避和自动资源创建等针对性恢复策略，确保您的 AI 智能体即使在不稳定环境中也能保持高效。

该技能对于构建长期运行或任务关键型自动化流水线的开发人员尤为重要。无论您是运行单个复杂函数还是编排整套通宵批量任务，自主执行器都能提供必要的韧性，助您将脆弱的脚本升级为使用 Openclaw Skills 的健壮生产级智能体工作流。

下载入口:https://github.com/openclaw/skills/tree/main/skills/ishsharm0/autonomous-executor

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

自主执行器 应用场景

• 自动化需要多步验证和错误处理的项目部署。
• 执行可能遇到速率限制或连接超时的深度网络研究任务。
• 在无需人工监管的情况下，安排并运行通宵批量处理作业。
• 构建在间歇性停机期间需要自我修正的弹性 API 集成。

1. 用户或智能体通过 Openclaw Skills 生态系统中的 AutoExec 或 Overnight API 组件触发任务。
2. 执行器分析环境并在可用工具中检查所需的能力。
3. 在执行期间，系统监控异常并将其分类为网络、认证或依赖项等特定类别。
4. 如果发生错误，引擎会应用相应的恢复策略，例如等待重试标头或自动安装缺失的库。
5. 执行进度将保存到 MongoDB 中的检查点，以便在发生终结性故障时可以恢复。
6. 生成最终报告，记录每次尝试、恢复操作和最终结果以供审核。

自主执行器 配置指南

使用 pip 安装软件包或将其包含在您的项目环境中。确保 MongoDB 实例可用于日志记录和检查点。

pip install autonomous-executor

配置您的环境变量以指向您的数据库：

export MONGODB_URI="mongodb://localhost:27017"

自主执行器 数据架构与分类体系

该技能利用 MongoDB 维持持久性，并为 Openclaw Skills 管理的自主操作提供可观测性。

Autonomous Executor

Self-healing, error-recovering task execution with full autonomy.

Features

• Automatic Error Recovery: Categorizes errors and applies targeted recovery strategies
• Exponential Backoff: Smart retry timing with configurable limits
• Self-Healing: Auto-installs missing dependencies, creates missing resources
• Checkpoint/Resume: Save progress for resumable overnight tasks
• MongoDB Logging: Persistent execution logs for monitoring
• Capability Awareness: Full knowledge of all OpenClaw skills and tools
• Overnight Batch Processing: Queue tasks to run while you sleep

Error Categories & Recovery

Quick Usage

from autonomous_executor.executor import AutoExec

# Deploy a project autonomously
result = AutoExec.deploy("my-portfolio", "portfolio", "My awesome portfolio")

# Run any function with auto-recovery
result = AutoExec.run(my_function, arg1, arg2, _max_retries=5)

# Check overnight report
report = AutoExec.report()

# List all capabilities
print(AutoExec.caps())

Overnight Batch Execution

from autonomous_executor.overnight import Overnight

# Queue tasks to run overnight
Overnight.deploy("my-portfolio", "portfolio", description="Overnight deploy")
Overnight.deploy("client-site", "landing", description="Client landing page")
Overnight.research("AI trends 2025")

# Queue custom functions
Overnight.run(my_backup_function, "/data", _name="Backup Data")

# Start overnight run
summary = Overnight.start()

# Next morning, check report
print(Overnight.report())

Decorator Usage

from autonomous_executor.executor import autonomous

@autonomous(max_retries=5, task_name="Critical Task")
def my_critical_function():
    # This will auto-recover from errors
    return do_something()

Full API

AutonomousExecutor

from autonomous_executor.executor import AutonomousExecutor

executor = AutonomousExecutor(user="ishaan")

# Execute any function
result = executor.execute(
    task_fn=my_function,
    args=(arg1, arg2),
    kwargs={"key": "value"},
    task_name="My Task",
    max_retries=5,
    timeout=300  # 5 minutes
)

# Run project pipeline
result = executor.run_project_pipeline({
    "name": "my-app",
    "type": "nextjs",
    "description": "My Next.js app",
    "extra": {"features": ["dark_mode", "auth"]}
})

# Run research
result = executor.run_research_pipeline("quantum computing", depth="deep")

# Get execution history
history = executor.get_execution_history(limit=20)

# Get overnight report
report = executor.get_overnight_report()

Example Output

[14:23:45] ?? Starting task: Deploy Portfolio
[14:23:45]    Max retries: 5, Timeout: 0s
[14:23:45]    Available capabilities: 15 skills
[14:23:45] ? Attempt 1/5...
[14:23:47] ? Attempt 1 failed: ConnectionError: Network timeout
[14:23:47]    Error category: network
[14:23:47] ?? Trying recovery: wait_and_retry
[14:23:47] ? Waiting 4.0s before retry...
[14:23:51] ? Attempt 2/5...
[14:24:03] ? Task completed successfully in 12.1s

MongoDB Collections

• task_records: Complete task execution records
• task_checkpoints: Checkpoint data for resumable tasks
• execution_logs: Real-time execution logs