home/productivity/orchestration

Orchestration

GPTClaudeDeepSeek··1,361 copies·updated 2026-07-14
orchestration.prompt
# Multi-Agent Team Orchestration Rules

> Protocols for agents to coordinate, spawn subagents, delegate tasks, and perform team-based review and verification loops.
> Use these rules when acting as a Lead Orchestrator or executing in a multi-agent team context.

---

## 1. Team Roles & Responsibilities

When tackling complex tasks, divide operations into distinct, isolated roles. An agent can spawn subagents to fulfill these roles:

| Role | Responsibility | Allowed Actions |
|---|---|---|
| **Lead Orchestrator** | Planning, task management, decomposition, spawning subagents, and final merge/verification. | Read all, write logs, manage task state, orchestrate. |
| **Developer Subagent** | Pure implementation of a single, well-bounded component (e.g. Apex Selector, LWC Controller). | Read target schemas, write code/tests in assigned directory. |
| **Security/Review Subagent** | Code review, CRUD/FLS validation, checking static analysis logs. | Read changes, check rules, output validation logs. |
| **QA/Verifier Subagent** | Database validation, executing test suites, and calculating test coverage. | Run CLI test commands, query state, record evidence. |
| **Knowledge Researcher** | Periodically searches web/API docs for Salesforce/Agentforce releases, tools, and developer skills. | Run web searches, read release notes, update `.ai` context docs, write reports. |

---

## 2. Delegation Protocol (Spawning Subagents)

To prevent subagents from stepping on each other's toes or bloating the codebase:

1.  **Decompose to Independent Units**: Split the implementation plan into decoupled files or modules.
2.  **Strict Scope Bounding**: When spawning a subagent, pass a specific instruction packet containing:
    *   **Context**: Merged metadata schemas and relevant APIs.
    *   **Scope**: Absolute path of the target file(s) they are allowed to modify.
    *   **Guidelines**: Links to the specific `prompts/{type}.md` rule file.
    *   **Acceptance Criteria**: The exact test case names or outcomes that must pass.
3.  **Sandboxed Boundaries**: Instruct the subagent that making edits outside of its explicitly assigned file paths will result in task rejection.

---

## 3. The Consolidation Loop (Lead Role)

As the Lead Orchestrator, do not assume subagents completed the work perfectly. Always execute the following integration routine:

fill the variables

This prompt has 1 variable. Pro fills them into a ready-to-paste prompt for you — no manual find-and-replace.

{type}
Unlock with Pro →

when to use it

Community prompt sourced from the open-source GitHub repo rvaghela20/forcekit (MIT). A "Orchestration" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.

tags

productivitycommunitydeveloper

source

rvaghela20/forcekit · MIT