skills.rest
guardiatechnology
guardiatechnologyOFFICIAL
Explore all skills by this author →
WebsiteGitHub

kata-agent-tools-design

Extract tools from orchestrator and specialist notes into deterministic, ML, and MCP categories.

Updated Sep 3, 2025
Search Tags:#MCP#schema validation#idempotency#guardrails#agent tools#tool catalog
One-click install
npx skills add https://github.com/guardiatechnology/design-system --skill kata-agent-tools-design
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: kata-agent-tools-design
Source: https://github.com/guardiatechnology/design-system/tree/main/.claude/skills/kata-agent-tools-design
Command: npx skills add https://github.com/guardiatechnology/design-system --skill kata-agent-tools-design

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps engineering teams precisely design an agent’s tool catalog by extracting tools, classifying them into deterministic/ML/MCP categories, and writing unambiguous runtime contracts with strong validation and idempotency guarantees.

Core Features & Use Cases

  • Tripartite tool catalog design: separates tools into Deterministic, ML, and MCP categories with explicit rules (no fourth category without ADR).
  • Contract-first documentation: generates a canonical tools.md template including I/O schemas, idempotency, lateral effects, latency, errors, and auth expectations for MCP.
  • Safety guardrails for runtime usage: enforces input validation at the tool boundary, prevents tenant boundary crossing, and blocks tool injection via cross-links to guardrails.

Quick Start

Ask an AI to generate docs/{context}/agents/{agent}/tools.md by extracting the full tool list from orchestrator.md and specialists/{name}.md, then classifying each tool into deterministic, ML, or MCP with complete input/output contracts and idempotency requirements.

Frequently Asked Questions about kata-agent-tools-design

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

FAQPage Schema
How do I design an agent tool catalog with schema validation and idempotency?▼

Design an agent tool catalog by extracting tools from orchestrator and specialist notes, then classifying them into deterministic, ML, and MCP categories. Generate a canonical tools.md file enforcing schema validation, idempotency for lateral-effect tools, and tenant-safe boundaries.

What is the best way to organize MCP tools to prevent tool injection in agent workflows?▼

Organize MCP tools into a tripartite catalog separating them from deterministic and ML tools. Apply safety guardrails by enforcing input validation at the tool boundary and cross-linking to guardrail-aligned protections to block tool injection.

How do I generate runtime contracts for an agent's tool catalog?▼

Generate runtime contracts by writing a canonical tools.md template that documents I/O schemas, idempotency requirements, lateral effects, latency, errors, and auth expectations for MCP tools extracted from orchestrator and specialist notes.

Why separate deterministic, ML, and MCP tools into different categories in an agent catalog?▼

Separating tools into deterministic, ML, and MCP categories enforces explicit operational rules and contract clarity. This tripartite design prevents category sprawl without an architecture decision record and ensures specific idempotency guarantees for lateral-effect tools.

Can I add a fourth custom tool category outside of deterministic, ML, and MCP?▼

No, the agent tool catalog design restricts classification to deterministic, ML, and MCP categories. Adding a fourth category requires an architecture decision record to justify the deviation from the established tripartite contract structure.

How do I enforce tenant-safe boundaries for MCP tools with lateral effects?▼

Enforce tenant-safe boundaries by applying input validation at the tool boundary and mandating idempotency for lateral-effect MCP tools. Document these constraints in tools.md alongside guardrail-aligned protections against tenant boundary crossing.

Related Skills

Explore More Software Engineering Skills→
vuralserhat86vuralserhat86
43

multi_agent_patterns

Design and implement multi-agent systems using architectural patterns and communication strategies.

Community
Advanced
friday-platformfriday-platform
97

debugging-runtime-errors

Map Friday agent runtime errors to one-line fixes.

Official
Intermediate
DektoraDektora

using-dekspec

Initializes DekSpec in your repository and manages its configuration.

Official
Advanced
89jobrien89jobrien
4

tool-presets

Standardize Claude Code agent tool access with predefined tool presets.

Community
Intermediate