agent-design_skill

This skill helps you configure agent profiles, define capabilities, and orchestrate complex workflows with subagents and commands.

TypeScript

0

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill hhopkins95/ai-systems --skill agent-design

SKILL.md9.6 KB

Overview

This skill helps define and orchestrate agent behavior, capabilities, and runtime integrations. It covers agent identity, system prompt, memory files, tools, skills, subagents, commands, and MCP server configuration. Use it to create reproducible agent profiles for complex multi-step workflows in TypeScript-based runtimes.

How this skill works

You define an AgentProfile object that declares identity, behavior, available tools, skills, subagents, commands, and MCP servers. The runtime reads that profile to grant tool access, load system prompts and memory files, spin up bundled MCP servers, and route command invocations to the correct flows. Skills and subagents encapsulate reusable processes and can be chained or invoked by commands to orchestrate multi-phase tasks.

When to use it

Creating or updating an agent's personality, system prompt, and project memory
Adding tools, skills, or subagents to extend agent capabilities
Defining commands that run multi-step workflows (plan → implement → test → review)
Bundling or referencing MCP servers for external integrations (GitHub, Notion, deploy)
Coordinating dependencies and environment variables for agent execution

Best practices

Keep systemPrompt concise for core behavior and use agentMDFile for project context and long-form memory
Model skills as focused, documented markdown files with supporting templates and dependency lists
Limit subagent tool access to the minimal subset needed for delegation
Expose high-level commands that orchestrate skills and subagents rather than embedding low-level steps everywhere
Bundle MCP servers for local integrations and list external MCPs for shared services

Example use cases

Build a 'code-assistant' profile that includes Read/Edit/Write tools, a code-review subagent, and an 'implement' command to automate feature delivery
Define an 'api-testing' skill with curl templates and a test-writer subagent to generate and run endpoint tests
Create a CI-style command that runs tests, builds, and deploys via a bundled deployment MCP server
Assemble a documentation sync command that fetches specs from a Notion MCP, updates markdown, and opens a PR via a GitHub MCP

FAQ

Use systemPrompt for core persona and hard constraints; store project-specific context and long-form memory in agentMDFile so context is maintainable and versionable.

When should I create a subagent versus a skill?

Create a subagent to delegate runtime tasks with distinct models or restricted tools. Use a skill to document reusable processes, templates, and dependency declarations.