agent-orchestration

Orchestrates multi-agent workflows to implement new features from free-text requirements.

Updated Apr 17, 2026
One-click install
npx skills add https://github.com/hpark0011/mirror --skill agent-orchestration-hpark0011
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-orchestration
Source: https://github.com/hpark0011/mirror/tree/main/.agents/skills/agent-orchestration
Command: npx skills add https://github.com/hpark0011/mirror --skill agent-orchestration-hpark0011

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Implementing a new feature across a monorepo requires coordinating exploration, planning, coding, validation, and quality gates, which is error-prone and token-expensive when done ad hoc. This Skill turns a free-text requirement into a structured multi-agent workflow with planned phases, paired executors and validators, and enforced quality gates. ## Core Features & Use Cases - Dynamic Phase Planning: Runs parallel Explorer and Planner agents, then validates the plan for ordering errors, parallel-safety, and scope creep before any code is written. - Scope-Based Execution: Chooses inline task spawning for small features, full agent teams for medium ones, and checkpointed teams for large ones, with per-phase executor and validator agents. - Quality Gates & Error Recovery: Runs type-check, lint, and build gates between phases with bounded retry loops, spot-check rules, and token budgeting to control cost. - Use Case: Ask to build a new comments feature in a packages/features directory; the Skill explores existing patterns, plans foundation through integration phases, spawns executor agents per phase, validates each output, and verifies final package exports with a full build. ## Quick Start Ask the agent to orchestrate the implementation of a new feature, for example: build a reactions feature in the features package using agents.

Frequently Asked Questions about agent-orchestration

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

FAQPage Schema
How do I implement a new feature using multiple AI agents?▼

Invoke the orchestration workflow with a free-text requirement. It runs Explorer and Planner agents to build a phased plan, validates the plan, then spawns executor and validator agents per phase with quality gates between them.

How does multi-agent orchestration control token costs?▼

It applies budgeting rules: no agent spawns for investigations under five file reads, one to two reviewers per phase, tight executor prompts under 800 tokens, and direct diff spot-checks instead of reviewer re-spawns for small fixes.

When should I not use agent orchestration for a coding task?▼

Skip orchestration for refactors, migrations, bug fixes, and documentation-only tasks; those are handled directly. Also prefer the spec-execution skill when a finished spec already exists, since planning phases would be redundant.

What happens when a validator agent rejects an executor's work?▼

The orchestrator spawns a retry executor with the validator's specific file and line feedback, up to two retries. If still rejected, it reports the failure to the user with options to fix, skip, or stop.

What quality gates run between implementation phases?▼

Gates include type-check, lint, combined type-check-and-lint, full build, and app compile verification, run via pnpm filter commands against the target package. A failed gate triggers one fix-agent retry before escalating to the user.