skills.rest
Oruga420
Oruga420COMMUNITY
Explore all skills by this author →
Website𝕏@Oruga420GitHub

long-running-harness

Orchestrate multi-agent planning, generation, and evaluation loops with sprint contracts and file-based handoffs.

Updated Mar 24, 2026
Search Tags:#multi-agent#generator#sprints#planner#harness#evaluator#context-reset
One-click install
npx skills add https://github.com/Oruga420/claude-code-skills --skill long-running-harness-oruga420
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: long-running-harness
Source: https://github.com/Oruga420/claude-code-skills/tree/main/long-running-harness
Command: npx skills add https://github.com/Oruga420/claude-code-skills --skill long-running-harness-oruga420

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Long-running projects and complex application builds lose coherence in single-agent sessions and suffer from flaky, low-quality outputs; this harness provides a structured multi-agent workflow to maintain quality over many hours or sprints.

Core Features & Use Cases

  • Three-agent architecture (Planner → Generator → Evaluator) with clear responsibilities to produce ambitious specs, incremental implementation, and independent, calibrated QA.
  • Sprint contracts and explicit acceptance criteria to negotiate scope and define testable "done" conditions before coding begins.
  • File-based handoffs and context resets to avoid context drift and enable fresh agents to continue work reliably across sprints.
  • Calibrated evaluator and scoring rubric to enforce hard quality thresholds and produce actionable fail/rework guidance.
  • Use Case: Running multi-sprint feature development (full-stack features, UI systems, game mechanics) where sustained, verifiable agent work and iterative evaluation are required.

Quick Start

Use the long-running-harness skill to run a Planner→Generator→Evaluator loop with sprint contracts, file-based handoffs, and context resets for a multi-sprint project.

Frequently Asked Questions about long-running-harness

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I maintain quality in multi-agent application builds across multiple sprints?▼

Multi-agent application builds maintain quality by orchestrating a Planner, Generator, and Evaluator loop with sprint contracts and file-based handoffs. This structured workflow prevents context drift and ensures incremental implementation meets calibrated quality thresholds across sprints.

What is the best way to prevent context drift in long-running AI agent projects?▼

Preventing context drift in long-running AI agent projects requires file-based handoffs and context resets between sprints. This approach allows fresh agents to continue work reliably without losing coherence over many hours of sustained development.

Can I use a multi-agent harness for full-stack feature development and UI systems?▼

Yes, a multi-agent harness supports full-stack feature development and UI systems by applying sustained agent work across multiple sprints. It negotiates scope with sprint contracts and defines testable acceptance criteria before coding begins.

How does an independent evaluator enforce quality thresholds in multi-agent workflows?▼

An independent evaluator enforces quality thresholds in multi-agent workflows by applying a calibrated scoring rubric. It produces actionable fail or rework guidance, ensuring generated code meets explicit acceptance criteria before sprint completion.

When should I use a multi-sprint planning and evaluation loop instead of a single-agent session?▼

Use a multi-sprint planning and evaluation loop instead of a single-agent session when building complex applications that require sustained, verifiable agent work. Single-agent sessions often lose coherence and produce flaky, low-quality outputs over long durations.

Does the long-running-harness skill require specific dependencies to orchestrate agents?▼

No, the long-running-harness skill does not require specific dependencies to orchestrate agents. It provides a self-contained architecture for planning, generation, and evaluation loops using file-based handoffs and context resets.

Related Skills

Explore More Software Engineering Skills→
X-School-AcademyX-School-Academy
9

dev-swarm-mvp

Define MVP scope, success metrics, and acceptance criteria for early product validation.

Official
Basic
steadyfallsteadyfall

canary

Monitor live applications for console errors and performance regressions after deployment.

Community
Advanced
cexllcexll
2.7k

codeagent

Orchestrate multi-backend AI code tasks with pluggable backends and file references.

Community
Advanced
ai-ecoverseai-ecoverse
28

delegation

Delegate parallel tasks to isolated sub-agents for efficient orchestration.

Official
Intermediate