Tool Definition Writer for Agents
Writes a clear, well-scoped tool definition (name, description, parameters, and usage guidance) for an AI agent, designed to minimize misuse and ambiguous tool-calling.
Developers building AI agents with custom tools (via function calling, MCP, or agent frameworks) who need tool definitions precise enough that the agent reliably picks the right tool at the right time.
You are an expert in designing tool definitions for AI agents that minimize ambiguous or incorrect tool calls.
What the tool does (in plain language): {{tool_functionality_description}}
Inputs the tool needs: {{required_inputs}}
What it returns: {{tool_output}}
Other tools the agent has access to (to avoid overlap/confusion): {{other_available_tools}}
Do the following:
1. Write a clear, specific tool name and a description written from the agent's perspective, explaining exactly when to use this tool versus when not to, especially distinguishing it from any similar tools listed.
2. Define the parameters needed, with a clear description for each (required vs optional, expected format, and any constraints like max length or valid value ranges).
3. Write 2-3 example scenarios showing when the agent should call this tool, and 1-2 counter-examples showing a similar-looking request where it should NOT call this tool (and what it should do instead).
4. Flag any part of the tool definition that is still ambiguous or could lead to misuse, and suggest how to tighten it further.How to use
- Describe the tool's functionality, required inputs, and output in plain language.
- List any other tools the agent has access to so the definition clearly distinguishes overlapping use cases.
- Copy the finalized name, description, and parameter definitions directly into your agent's tool configuration.
- Address any flagged ambiguity before deploying, since unclear tool definitions are a common source of agent misbehavior.
Example input
Tool functionality: Sends a Slack message to a specified channel. Inputs: channel name, message text. Output: confirmation of send success/failure. Other tools: 'create_calendar_event', 'send_email'.
Example output
A tool named 'send_slack_message' with a description clarifying it's for internal team notifications only, not customer-facing emails (distinguishing it from 'send_email'), parameter definitions for channel (required, must match an existing channel name) and message (required, max 4000 characters), example scenarios like notifying a team about a completed task, a counter-example of a customer follow-up request that should use 'send_email' instead, and a flag that the tool definition doesn't specify what happens if the channel doesn't exist, with a suggestion to add explicit error-handling guidance.
Customization tips
- — If building for a specific framework (OpenAI function calling, Anthropic tool use, MCP), mention it so the output format matches that framework's schema conventions.
- — For tools with side effects (sending messages, making purchases, deleting data), explicitly ask for extra cautionary language in the description to reduce accidental misuse.
- — Request a JSON schema version of the parameters if you need it directly usable in code.
Tags
FAQ
What is this prompt for?
How should I customize it?
Are there any limitations?
How is it different from a basic prompt?
Related Prompts
All AI Agents prompts →Agent System Prompt Stress Tester
Pressure-tests an AI agent's system prompt by generating edge-case user inputs designed to expose ambiguity, scope creep, or unsafe behavior before deployment.
Agent Task Decomposition Planner
Breaks a broad goal into a structured set of sub-tasks an AI agent can execute step by step, with clear success criteria and handoff points for human review.
Agent Failure Post-Mortem Analyzer
Analyzes a transcript or log of an AI agent that failed, went off-track, or produced a bad output, and identifies the root cause with a specific prompt or logic fix.
Related Skills
Agent Prompt & Tool Spec Designer
Designs a complete system prompt and tool specification for an LLM agent from a description of what the agent should do.
MCP Server Security Reviewer
Static, read-only security review for MCP servers and tool handlers — checks for prompt injection surfaces, secrets leakage, and unsafe tool permissions before deployment.
AI Agent Prompt & Tool Spec Builder
Designs a complete AI agent specification — system prompt, tool definitions, and decision boundaries — from a description of the task you want the agent to handle.
Related Articles
How do I add memory to an AI agent or n8n workflow?
Memory can mean chat history, user preferences, retrieved documents, or stored workflow state. It should be scoped, editable, and limited to what the task needs.
How do I evaluate an AI agent before trusting it?
Evaluate an AI agent with realistic tasks, edge cases, tool failures, permission checks, output rubrics, logs, and human review before using it in production.
What should an AI agent never be allowed to do without approval?
An AI agent should not send messages, spend money, delete files, publish content, change records, access secrets, or contact customers without clear approval rules.