manager

Parse planning contracts and enforce guardrails across multi-agent loops.

Updated Mar 19, 2026

Search Tags:#multi-agent #coordination #evidence #guardrails #manager #planning-contract

Ready to run

npx skills add https://github.com/espensev/ai-skills --skill manager-espensev

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: manager
Source: https://github.com/espensev/ai-skills/tree/main/gemini-skills/skills/manager
Command: npx skills add https://github.com/espensev/ai-skills --skill manager-espensev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Orchestrates multiple agents by interpreting a centralized planning contract and enforcing global guardrails to prevent unsafe writes and inconsistent states.

Core Features & Use Cases

Parses the planning contract and coordinates agent actions across the loop.
Enforces scoped writes and maintains evidence trail for verification.
Supports adaptive coordination across planning, execution, QA, and reporting phases.

Quick Start

Instruct the Manager Agent to start coordinating the multi-agent loop by parsing the current planning contract.

Frequently Asked Questions about manager

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

FAQPage Schema

How do I coordinate multi-agent loops to prevent unsafe writes?▼

Multi-agent coordination uses a centralized planning contract to enforce global guardrails and orchestrate agent actions. This prevents unsafe writes and inconsistent states across planning, execution, QA, and reporting phases while maintaining an evidence trail for verification.

What is a planning contract in multi-agent orchestration?▼

A planning contract is a centralized document, specifically docs/planning-contract.md, that defines scoped changes and constraints for multi-agent orchestration. The manager agent parses this contract to enforce guardrails and direct execution across planning, execution, QA, and reporting phases.

How do I enforce guardrails for multi-agent execution and reporting?▼

You enforce guardrails by constraining multi-agent outputs to scoped changes and maintaining an evidence trail. The manager interprets the planning contract to restrict unsafe writes, ensuring disciplined and auditable modifications across all coordinated execution and reporting phases.

Can I use a planning contract to manage adaptive coordination across phases?▼

Yes, a planning contract supports adaptive coordination across planning, execution, QA, and reporting phases. The manager agent parses the contract and recent QA findings to dynamically coordinate multi-agent loops and constrain outputs to scoped changes.

What inputs do I need to start coordinating a multi-agent planning loop?▼

You need a centralized planning contract, modified files, recent QA findings, and an output contract. The manager agent parses these inputs to enforce global guardrails, coordinate multi-agent actions, and produce auditable evidence trails for scoped changes.

When should I not use centralized planning contracts for agent coordination?▼

Centralized planning contracts are not suited for unstructured or isolated agent tasks lacking scoped changes or QA findings. Without the need for an auditable evidence trail or global guardrails across planning and execution, this orchestration approach adds unnecessary overhead.