orchestrate-implementation

Executes spec-driven implementation plans with orchestrated executor and verifier agent waves.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Multi-file feature implementation often burns tokens on unnecessary agent spawns, redundant reviews, and mid-flight clarification rounds. This Skill executes a finished spec's Team Orchestration Plan with disciplined wave-based execution, keeping orchestrator, executor, and verifier roles separate and costs under control. ## Core Features & Use Cases - Wave-based execution: Groups spec steps into waves scoped by verification gates, with one executor agent per wave and one commit per wave as a rollback checkpoint. - Budget-disciplined verification: Dispatches a minimal set of parallel code-review verifiers per wave using a routing table, and spot-checks small deterministic fixes instead of re-spawning reviewers. - Feedback loops to green: Consolidates verifier findings by severity and iterates with the executor in its existing thread until all Critical findings are resolved. - Use Case: You have a finished spec at workspace/spec/chat-constraints-spec.md and want it implemented across backend, frontend, and tests with verified gates at each stage, without wasting tokens on ad-hoc agent orchestration. ## Quick Start Implement the spec at workspace/spec/chat-constraints-spec.md by running its orchestration plan wave by wave.

Frequently Asked Questions about orchestrate-implementation

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

FAQPage Schema
How do I implement a spec with multiple agents?▼

Read the spec end-to-end, group its steps into waves scoped by verification gates, then spawn one executor agent per wave followed by one or two parallel verifier agents on the diff. Commit each wave separately and iterate on verifier feedback until clean.

When should I use orchestrate-implementation vs agent-orchestration?▼

Use orchestrate-implementation when a finished spec exists at workspace/spec with a Team Orchestration Plan. Use agent-orchestration for free-text requirements without a spec, since it includes its own exploration and planning phase.

How many code review agents should review each wave?▼

Default to one or two reviewers per wave, adding a third only when work spans correctness plus a specialist axis like concurrency or data integrity. Never add reviewers for completeness, since each reviewer costs 30-50k tokens.

Should I re-run all reviewers after a small fix?▼

No. For deterministic fixes touching fewer than three files, spot-check the diff yourself and re-run the hard gates. Re-dispatch only the affected reviewers when new logic or tests are added, and reserve full re-review for cross-cutting refactors.

When should I not use spec orchestration for implementation?▼

Skip it for single-file edits or one-shot bug fixes, since agent spawn overhead of 10-20k tokens exceeds the value of a small change. Also avoid it for pure research or for authoring the spec itself, which belongs to create-spec.