Agent Loop Hub

English | 中文

Agent Loop Hub is a community collection of reusable AI Agent workflow loops.

Description

Agent Loop Hub collects practical, reusable workflow templates that show how AI agents plan, execute, review, recover from failure, and hand control back to humans. Instead of sharing a single prompt, each loop describes the roles, state, checkpoints, and repeatable process behind an agent-powered workflow.

Most people share prompts. This repo shares loops.

A loop defines how agents repeatedly plan, act, review, fix, and hand control back to humans when needed.

People share prompts. We share loops.

What Is An Agent Loop?

An Agent Loop is a reusable workflow pattern for AI agents.

It answers:

• Who plans?
• Who executes?
• Who reviews?
• When should the human step in?
• Where is state stored?
• How does the workflow continue after failure?

Loop Gallery

Category	Loop	Use Case	Status
Coding	OpenSpec Dev Loop	Plan -> Spec -> Code -> Review -> Test -> Archive	Ready
Coding	Superpowers Dev Loop	Brainstorm -> Plan -> Subagent TDD -> Review -> Finish	Ready
Research	Deep Research Loop	Search -> Read -> Synthesize -> Source Check -> Report	Planned
Writing	Content Creator Loop	Research -> Outline -> Draft -> Critique -> Revise	Planned
Productivity	Meeting Assistant Loop	Transcript -> Summary -> Action Items -> Follow-up	Planned
Business	Lead Scoring Loop	Collect -> Enrich -> Score -> Human Review -> CRM Update	Planned
Data	Data Analysis Loop	Load -> Clean -> Analyze -> Visualize -> Report	Planned

Each loop is tagged with a source type in its loop.yaml:

Type	Meaning
original	Designed from scratch by Agent Loop Hub contributors.
inspired_by	Conceptually inspired by another project, but all prompts, templates, and structure are original.
adapted_from	Based on someone else's template or workflow, restructured into the Agent Loop Hub format.

All source URLs and licenses are documented per-loop. See each loop's README.md and loop.yaml for details, or refer to ATTRIBUTIONS.md for a full summary.

Repository Structure

agent-loop-hub/
  README.md
  README.zh-CN.md
  loops/
    coding/
      openspec-dev-loop/
        README.md
        loop.yaml
        template/
          ORCHESTRATOR.md
          state.template.md
          prompts/
            planner.prompt.md
            executor.prompt.md
            reviewer.prompt.md
      superpowers-dev-loop/
        README.md
        loop.yaml
        template/
          ORCHESTRATOR.md
          state.template.md
          prompts/
            brainstorming.prompt.md
            writing-plans.prompt.md
            subagent-driven-development.prompt.md
            requesting-code-review.prompt.md
            finishing-a-development-branch.prompt.md
    research/
      deep-research-loop/
        README.md
        loop.yaml
        template/
          ORCHESTRATOR.md
          state.template.md
          prompts/
    writing/
      content-creator-loop/
        README.md
        loop.yaml
        template/
          ORCHESTRATOR.md
          state.template.md
          prompts/
    productivity/
      meeting-assistant-loop/
        README.md
        loop.yaml
        template/
          ORCHESTRATOR.md
          state.template.md
          prompts/
    business/
      lead-scoring-loop/
        README.md
        loop.yaml
        template/
          ORCHESTRATOR.md
          state.template.md
          prompts/
    data/
      data-analysis-loop/
        README.md
        loop.yaml
        template/
          ORCHESTRATOR.md
          state.template.md
          prompts/
  schemas/
    loop.schema.yaml
  ATTRIBUTIONS.md
  CONTRIBUTING.md
  LICENSE

What Belongs Here?

This repository is for complete agent workflow loops, not one-off prompts.

A good loop should include:

• Role definitions
• Execution flow
• Human checkpoints
• State persistence rules
• Failure and retry behavior
• Prompt templates for each role

License

MIT