grace-multiagent-execute

Coordinate batched GRACE development plan execution across multiple AI agents with controller-managed parallel waves.

233|51|Updated Feb 17, 2026
One-click install
npx skills add https://github.com/osovv/grace-marketplace --skill grace-multiagent-execute
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: grace-multiagent-execute
Source: https://github.com/osovv/grace-marketplace/tree/main/skills/grace/grace-multiagent-execute
Command: npx skills add https://github.com/osovv/grace-marketplace --skill grace-multiagent-execute

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Coordinating multiple AI workers to implement different modules in parallel often leads to inconsistent shared artifacts, architectural drift, and merge conflicts; this Skill enforces controller-owned planning, scoped worker packets, and batched graph synchronization to keep the plan and knowledge graph consistent.

Core Features & Use Cases

  • Build parallel-safe execution waves from docs/development-plan.xml and docs/knowledge-graph.xml and present the proposed waves for user approval.
  • Produce compact execution packets per module that include contract excerpts, write scopes, verification commands, and expected graph delta fields.
  • Enforce explicit ownership: controller for shared artifacts and workers for module-local code, with scoped reviewers validating outputs before controller integration.
  • Support selectable safety profiles (safe, balanced, fast), batched graph sync, wave-level integration checks, and controller-only commits for shared artifacts.
  • Use Case: accelerate implementation across many independent modules while maintaining a single source of architectural truth and automated verification gating.

Quick Start

Run a balanced multi-agent execution of the development plan with controller-managed waves and targeted graph sync.

Frequently Asked Questions about grace-multiagent-execute

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

FAQPage Schema
How do I parallelize multi-agent code execution without causing merge conflicts?

You coordinate multi-agent execution by batching module work into controller-managed parallel waves with scoped worker packets. A controller agent owns shared artifacts and commits, while fresh per-module workers implement local code changes and reviewers validate scoped graph deltas before integration.

What is controller-managed parallel execution for AI agents?

Controller-managed parallel execution is a coordination pattern where a central controller agent builds execution waves, assigns scoped packets to worker agents, and reviews graph deltas before integrating shared artifacts. This prevents architectural drift when multiple AI workers implement different modules simultaneously.

How do I execute a development plan across multiple agents in parallel?

To execute a development plan across multiple agents, you parse docs/development-plan.xml and docs/knowledge-graph.xml to build parallel-safe execution waves. The controller presents proposed waves for user approval, then dispatches compact execution packets containing contract excerpts and verification commands to module-local workers.

Do I need a knowledge graph and development plan to coordinate multiple AI workers?

Yes, you need docs/development-plan.xml and docs/knowledge-graph.xml present to coordinate multiple AI workers. These files provide the structured plan and architectural truth required to build parallel-safe execution waves, define write scopes, and track scoped graph deltas during batched wave executions.

Can I select different safety profiles for parallel module execution?

Yes, you can select from safe, balanced, and fast safety profiles for parallel module execution. These profiles control the strictness of batched graph synchronization, wave-level integration checks, and verification gating applied by the controller during multi-agent wave execution.

What are the limitations of batched wave execution for multi-agent development?

Batched wave execution requires strict controller ownership of shared artifacts and cannot allow workers to commit directly to shared paths. Limitations include the overhead of centralized graph synchronization and the necessity of fresh per-module worker agents to prevent state leakage between waves.