tool-design

Designs agent-facing tool descriptions, schemas, error messages, and consolidated tool catalogs.

Updated Aug 7, 2026
One-click install
npx skills add https://github.com/Sambhav242005/Major-Project --skill tool-design-sambhav242005
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: tool-design
Source: https://github.com/Sambhav242005/Major-Project/tree/main/.agents/skills/tool-design
Command: npx skills add https://github.com/Sambhav242005/Major-Project --skill tool-design-sambhav242005

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Agents pick the wrong tool or generate malformed calls when tool descriptions are vague, overlapping, or inconsistent. This Skill provides a systematic method for writing tool descriptions agents can route on, designing schemas and response formats, and consolidating bloated tool catalogs. ## Core Features & Use Cases - Description Engineering: Structure every tool description to answer what the tool does, when to use it, what inputs it accepts, and what it returns, with concrete format examples. - Catalog Consolidation: Apply the consolidation principle to merge overlapping tools, use verb-noun naming and MCP namespacing (ServerName:tool_name), and decide when architectural reduction to primitive tools outperforms specialized ones. - Actionable Error Design: Generate structured error messages that tell agents what went wrong, the expected format, and how to retry. - Use Case: An agent with 17 overlapping tools picks the wrong one half the time. Use this Skill to audit the catalog with the included checklist, consolidate redundant tools, rewrite descriptions, and validate quality with the scoring evaluator script. ## Quick Start Ask the agent to audit your tool catalog and rewrite each tool's description, schema, and error messages following the tool-design checklist.

Frequently Asked Questions about tool-design

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

FAQPage Schema
How do I write a tool description an agent can route on?

Structure the description to answer four questions: what the tool does, when to use it, what inputs it accepts with types and format examples, and what it returns including error conditions. Avoid vague phrases like "helps with" and include concrete examples such as "CUST-000001" for ID formats.

How do I fix an agent picking the wrong tool from a large catalog?

Apply the consolidation principle: merge tools with overlapping activation scenarios until each has one unambiguous purpose, and namespace the rest by domain (e.g., db_*, web_*). If a human engineer cannot say which tool applies, the agent cannot either.

When should I consolidate tools versus keep them separate?

Consolidate tools that are sequential steps of one workflow into a single comprehensive tool. Keep tools separate when they have fundamentally different behaviors, serve different contexts, or must be callable independently; over-consolidation past 8-10 parameters hurts parameterization accuracy.

How should MCP tool names be formatted to avoid tool-not-found errors?

Always use fully qualified names in the format ServerName:tool_name, such as BigQuery:bigquery_schema. Without the server prefix, agents may fail to locate tools when multiple MCP servers expose similarly named tools.

What makes an error message recoverable for an agent?

A recoverable error states what went wrong, the expected format with a concrete example, and what to change before retrying, plus whether the error is retryable. Messages like "failed" or "invalid input" give the agent no recovery signal.

When does reducing tools to primitives outperform specialized tools?

Reduction works when the data layer is well-documented, the model has sufficient reasoning capability, and specialized tools constrain rather than enable the model. Avoid it when data is messy, domain expertise is required, or safety constraints demand restricted actions.