What problem does it solve?
Manually coordinating multiple AI agents for iterative project improvements, tracking progress, and verifying results is time-consuming and prone to human error. This Skill eliminates that overhead by automating the entire orchestration, verification, and insight accumulation workflow for any project with pre-defined agents.
Core Features & Use Cases
- Dynamic Agent Orchestration: Auto-detects and classifies all agents defined in your project's .claude/agents/ directory with no hardcoded agent names, works for any project structure.
- Mega Loop Iterative Improvement: Runs N-round (default 20) iterative improvement loops with automatic issue carryover between rounds, progress tracking, and final comprehensive reporting.
- Built-in Quality Gates: Includes mandatory Playwright browser verification for web apps at the end of each loop, plus automated insight accumulation across domain-specific teams.
- Task Router for One-off Workflows: Routes single tasks like audits, builds, or cleanups to the appropriate agents with verification and insight logging, no loop required.
- Use Case: For a web app project with defined frontend, backend, and QA agents, use this Skill to run a full 20-round improvement cycle, automatically assigning tasks, verifying each round's work, and generating a final report with remaining issues.
Quick Start
Use the harness-loop skill to run a 20-round iterative improvement loop for the current web app project, automatically coordinating all defined agents and verifying final results with Playwright browser checks.
