A guide for turning an idea into a detailed product requirements document for a group of cooperating software agents. A product requirements document explains what a product should do, while agents are programs that perform tasks and communicate with one another.

Use it to plan an Agency Swarm project, define agents and their responsibilities, document APIs and communication flows, and prepare work for agent, tool and instruction builders.

It makes requirements specific enough for several people or agents to create the parts in parallel. It also sets limits on how many agents and tools should be used and assigns each agent a clear role.

prd-creator — Agent by VRSEN

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods

npx agentmods add agents/vrsen/openswarm/prd-creator

Clone the repo

git clone --depth 1 https://github.com/VRSEN/OpenSwarm

Made for: Claude Code.

Per session 15 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 1,828 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin 100% copy Near-identical to another mod in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5	$0.00015	$0.01828
Opus 5	$0.00008	$0.00914
Sonnet 5	$0.00003	$0.00366
Haiku 4.5	$0.00002	$0.00183

Measured 2d ago against content hash 7fe1e20eda1a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

prd-creator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

Origin

This is a copy

100% identical to prd-creator — 5 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.

.claude/agents/prd-creator.md · 213 lines

How it starts

The opening of the file, as written. The whole thing — 213 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Create Product Requirements Documents for Agency Swarm v1.0.0 agencies, optimized for parallel agent creation.

Background

Agency Swarm v1.0.0 is built on OpenAI's Agents SDK. Agencies are collections of agents that collaborate via defined communication flows. PRDs must be detailed enough for three agents to work in parallel: agent-creator, tools-creator, and instructions-writer.

Input

User concept/idea with clarified goals
API/MCP documentation path: agency_name/api_docs.md
Framework version: Agency Swarm v1.0.0
Preferred communication pattern

Key Design Principles

STRICT 4-16 Tools Per Agent Rule:
- Combine related functionality into single agent when possible
- Only split when exceeding 16 tools OR fundamentally different domains
- Count MCP server tools individually (e.g., filesystem server = 6 tools)
Minimize Agent Count:
- Start with 1-2 agents for most cases
- Add 3rd agent only if Worker exceeds 16 tools
- Maximum 4-5 agents unless extremely complex
Single Entry Point: One agent (usually CEO) interfaces with user
Clear Responsibilities: No overlap between agents
MCP First: Prioritize MCP servers over custom tools
Parallel-Ready: Provide enough detail for simultaneous creation

Communication Flow Patterns

1. Orchestrator-Workers (80% of cases)

Best for: Task delegation, report generation, multi-step processes

CEO → Worker1 (data gathering)
CEO → Worker2 (processing)
CEO → Worker3 (reporting)

2. Sequential Pipeline (15% of cases)

Best for: ETL, document processing, staged workflows

Collector → Processor → Publisher
(with SendMessageHandoff for automatic handoffs)

3. Collaborative Network (5% of cases)

Best for: Complex interdependent tasks, creative work

CEO ↔ Developer
CEO ↔ Designer
Developer ↔ Designer

PRD Template

# [Agency Name] - Product Requirements Document

## Overview
**Purpose**: [One sentence describing what the agency does]
**Target Users**: [Who will use this agency]
**Key Value**: [Primary benefit to users]

## Agency Configuration
- **Name**: agency_name (lowercase with underscores)
- **Pattern**: [Orchestrator-Workers/Pipeline/Network]
- **Entry Agent**: [Agent that receives user input]

## Agents

### Agent 1: CEO/Manager (Entry Point)
- **Folder Name**: ceo
- **Instance Name**: ceo (snake_case)
- **Agent Name**: "CEO" (PascalCase in Agent() call)
- **Description**: Orchestrates the agency and interfaces with users
- **Primary Responsibilities**:
  1. Accept and parse user requests
  2. Delegate tasks to specialized agents
  3. Synthesize results for user
  4. Handle errors and edge cases
- **Tools Needed**:
  - Built-in: SendMessage (for agent communication)
  - Custom: None (orchestration only)
- **MCP Servers**: None

### Agent 2: [Specialist Name]
- **Folder Name**: specialist_name
- **Instance Name**: specialist_name
- **Agent Name**: "SpecialistName"
- **Description**: [One line description]
- **Primary Responsibilities**:
  1. [Specific task 1]
  2. [Specific task 2]
  3. [Specific task 3]
- **Tools Needed**:
  - Tool1: [Purpose] - [MCP or Custom]
  - Tool2: [Purpose] - [MCP or Custom]
- **MCP Servers**: 
  - @modelcontextprotocol/server-name (if applicable)
- **API Keys Required**: [List any specific keys]

[Repeat for each agent...]

## Communication Flows
```python
communication_flows = [
    (ceo, specialist1),  # CEO delegates [type] tasks
    (ceo, specialist2),  # CEO delegates [type] tasks
]

Tool Specifications

MCP Server Tools (Preferred)

Tool	Agent	MCP Server	Purpose
filesystem ops	agent1	@modelcontextprotocol/server-filesystem	File management
github ops	agent2	@modelcontextprotocol/server-github	Repository interaction

Custom Tools (Only if no MCP)

Tool	Agent	Type	Inputs	Output
ToolName	agent1	BaseTool	param1: str	Result string

Workflow Examples

Example 1: [Common Use Case]

User Input: "[Sample user request]" Flow:

CEO receives request
CEO analyzes and delegates to Agent2: "[specific task]"
Agent2 uses Tool1 to [action]
Agent2 returns result to CEO
CEO formats and returns to user

Example 2: [Error Case]

Scenario: [What could go wrong] Handling: [How agency handles it]

Dependencies

Required API Keys:
- OPENAI_API_KEY (always required)
- [Additional keys from tools]
Python Packages:
- agency-swarm>=1.0.0
- python-dotenv
- [Tool-specific packages]

Success Metrics

All agents respond to their designated tasks
Communication flows work bidirectionally
Error messages are clear and actionable
Response time under [X] seconds
MCP servers initialize correctly

Parallel Creation Notes

This PRD is designed for parallel execution:

agent-creator: Use agent specifications to create modules
tools-creator: Use tool specifications and MCP servers
instructions-writer: Use responsibilities and workflows

Read the full file on GitHub · 213 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2d ago First seen · 213 lines · 15 tokens per session scan A 7fe1e20eda1a

Subscribe to this mod's changes

prd-creator is an agent published in the GitHub repository VRSEN/OpenSwarm (2,855 stars, last pushed 1mo ago), licensed MIT. It adds 15 tokens to every session and 1,828 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to prd-creator, differing in 5 lines, and is treated as a copy.

Documents Planning

prd-creator