编程
Loop Engineering AI programming paradigm guide. Covers the core concepts of transitioning from single prompt calls to autonomous loop systems, key components, loop patterns, configuration examples
Loop Engineering AI programming paradigm guide. Covers the core concepts of transitioning from single prompt calls to autonomous loop systems, key components, loop patterns, configuration examples, and anti-patterns.
Loop Engineering is a new AI programming paradigm -- designing autonomous loop systems that allow Agents to continuously execute tasks in an environment with goals, feedback, and self-verification, eliminating the need for manual prompting.
Use when users need to learn about "Loop Engineering", "AI loop systems", "autonomous Agent design", "/loop", "/goal", "AI programming paradigms", or "system architect".
Transition from single prompt calls to continuous loop systems:
| Paradigm | How It Works | Developer Role |
|---|---|---|
| Prompt Engineering | Manually enter prompts, model executes once, human judges results | Prompt Engineer |
| Harness Engineering | Build a constrained environment for a single Agent with pre-checks/fixes/hooks, still requires human triggering | Harness Engineer |
| Loop Engineering | Manage multi-Agent sequencing, decision-making, and autonomous loops, self-driven over long periods across many rounds | System Architect / Loop Engineer |
| Component | Purpose | Reference |
|---|---|---|
| Automations | Trigger tasks by time/event, support periodic execution and archiving | See config-examples |
| Worktrees | Manage task state and context, maintain memory and state persistence across rounds | See config-examples |
| Skills | Encapsulate reusable task capabilities as skill modules for the loop system to invoke | See patterns |
| Plugins/Connectors | Extend system capabilities, interact with external tools or services | See integration |
| Sub-agents | Small specialized Agents for division of labor, supporting parallel multi-Agent task execution | See patterns / config-examples |
| External Memory & Feedback Loops | Achieve self-assessment and correction through mechanisms like logs, tests, and type checking | See anti-patterns |
| Primitive | Behavior | Use Case | Usage Tip |
|---|---|---|---|
/loop | Execute tasks repeatedly on a cycle | Scheduled checks, periodic maintenance | Must set maxRounds; no termination condition = infinite loop |
/goal | Run continuously until verifiable conditions are met | Goal-oriented autonomous tasks, with an independent model checking completion each round | Completion conditions must be objectively verifiable (file exists, tests pass, metrics met) |
/loop for the former, /goal for the latter.| File | Content | Who It's For |
|---|---|---|
| patterns.md | 8 common loop patterns + selection guide | Need to quickly find a matching pattern |
| config-examples.md | Automation / Worktree / Sub-agent configuration examples | Starting to build a loop system |
| anti-patterns.md | Common mistakes + correction strategies | Troubleshooting or preventing issues |
| integration.md | OpenClaw / TaskFlow and other platform integration | Connecting to actual toolchains |
Design the engineered loop for a medium/large (semi-)autonomous AI-coding task by decomposing it into gated sub-loops, emitted as a runnable .loop/ runbook. Use-when: "design an agent loop", "set up an autonomous / self-running agent workflow", "$loop-constructor". It DESIGNS the loop; it does NOT execute it.
Use when an AI coding agent needs bounded development loops, persistent project-local Docs/ state, context budgeting, environment escalation rules, safe stop...
Autonomous task loop engine — self-verifying, goal-driven, human-out-of-loop. Triggered by: loop, team-loop, 迭代研发, 迭代. Human sets a quantifiable goal (e.g.,...
Design an engineered gated loop for a medium/large (semi-)autonomous AI-coding task executed with the Codex CLI (single-agent, `codex exec`), emitted as a runnable .loop/ runbook. Use-when: "design an agent loop for codex", "$loop-constructor-codex". It DESIGNS the loop; it does NOT execute it.
Loops (loops.so). Use this skill for ANY Loops request — reading, creating, updating, and deleting data. Whenever a task involves Loops, use this skill inste...
Improve important deliverables by looping them through multiple isolated AI reviewers, each evaluating from a different angle, until all reviewers give full...