mcpField
L
loop-agent

loop-agent

github.com/d-o-hubAgent workflows1363 installsUpdated today
Repository

About loop-agent

A skill for running workflow agents iteratively to improve results until quality criteria are met. It supports fixed, criteria-based, convergence, and hybrid stopping modes for tasks such as code refinement, test-fix cycles, and performance optimization.

Key features

  1. 01Use four loop termination modes
  2. 02Apply common iterative workflow patterns
  3. 03Refine results against quality criteria
  4. 04Support test-fix and optimization cycles

Use cases

  1. 01Refine code over multiple passes to meet standards
  2. 02Repeat tests and fixes
  3. 03Improve performance progressively and assess gains

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/d-o-hub/rust-self-learning-memory --skill loop-agent

After installing, type/loop-agentin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

configuration.md
examples.md
modes.md
patterns.md
evals/
evals.json

Details

From the project README

Loop Agent Skill

Execute workflow agents iteratively for refinement and progressive improvement until quality criteria are met.

Quick Reference

  • Modes - Loop termination modes (fixed, criteria, convergence, hybrid)
  • Patterns - Common loop patterns (refinement, test-fix, optimization)
  • Configuration - Loop setup and templates
  • Examples - Complete loop examples

When to Use

  • Code needs iterative refinement until quality standards met
  • Tests need repeated fix-validate cycles
  • Performance requires progressive optimization
  • Quality improvements need multiple passes
  • Feedback loops necessary for convergence

NOT Appropriate For

  • Single-pass tasks (use specialized agent)
  • Purely parallel work (use agent-coordination)
  • Simple linear workflows (use sequential)
  • One-time analysis

Core Concepts

Loop Termination Modes
ModeDescriptionUse When
FixedRun exactly N iterationsKnown number of passes needed
CriteriaUntil success criteria metSpecific quality/performance targets
ConvergenceStop at diminishing returnsOptimal result unknown
HybridCombine multiple conditionsComplex requirements

See modes.md for detailed mode documentation and patterns.md for common loop patterns.

Comments

Markdown supported