agent-blueprint-skills

Agent Blueprint Skills

Open-source AI agent skills for enterprise deployment. Built for coding agents (Claude Code, Codex, Cursor, Copilot, Windsurf) using the Agent Skills open standard.

Four skills. Zero vendor lock-in. Works with any coding agent.

Skill	What it does
agent-fundamentals	Agentic AI fundamentals: what agents are, when to use them vs. workflows, architecture patterns, decision frameworks, failure modes, worked examples
servicenow-ai-agents	ServiceNow AI Agent Studio reference: data model, entity mapping, tool types, prompting patterns, browser navigation, MCP setup
agent-deployment	Vendor-agnostic deployment methodology: phased rollout, pilot-first, test gates, access tier adaptation
research-agent	Multi-tool research with automatic depth routing: free tools first, paid escalation, caching, structured output

Install

Claude Code

Add the marketplace, then install the plugin:

/plugin marketplace add agent-blueprint/agent-blueprint-skills
/plugin install agent-blueprint-skills@agent-blueprint

Skills appear in your session automatically. Claude loads them when your task matches the skill description.

OpenAI Codex CLI

Copy skills into your project:

git clone https://github.com/agent-blueprint/agent-blueprint-skills.git /tmp/ab-skills
cp -r /tmp/ab-skills/plugins/agent-blueprint-skills/skills/* .agents/skills/

Cursor / Copilot / Windsurf / Any Agent

Clone this repo and point your agent to the SKILL.md files:

git clone https://github.com/agent-blueprint/agent-blueprint-skills.git .agent-skills

The SKILL.md format is an open standard. Any agent that reads markdown can use these skills. Skills are in plugins/agent-blueprint-skills/skills/.

Manual (single skill)

Copy just the skill you need:

# ServiceNow AI Agents
curl -o SKILL.md https://raw.githubusercontent.com/agent-blueprint/agent-blueprint-skills/main/plugins/agent-blueprint-skills/skills/servicenow-ai-agents/SKILL.md

# Deployment Methodology
curl -o SKILL.md https://raw.githubusercontent.com/agent-blueprint/agent-blueprint-skills/main/plugins/agent-blueprint-skills/skills/agent-deployment/SKILL.md

# Research Agent
curl -o SKILL.md https://raw.githubusercontent.com/agent-blueprint/agent-blueprint-skills/main/plugins/agent-blueprint-skills/skills/research-agent/SKILL.md

What's in each skill

agent-fundamentals

A practical guide for understanding AI agents in enterprise environments:

• Agent vs. workflow vs. chatbot: When do you actually need an agent? Decision framework with clear criteria
• Anatomy of an agent: Instructions, tools, memory, guardrails, orchestration
• Architecture patterns: Single agent, orchestrator + workers, pipeline, human-in-the-loop — with selection criteria
• Common failure modes: Over-agenting, tool sprawl, vague instructions, missing guardrails, skipping validation
• Use case evaluation framework: 9-question walkthrough to determine if a process needs an agent
• Worked examples: Three detailed walkthroughs (simple ticket triage, workflow-not-agent invoice processing, complex multi-agent procurement)

servicenow-ai-agents

Everything a coding agent needs to deploy AI agents on ServiceNow's AI Agent Studio:

• Data model: All sn_aia_* tables, fields, and relationships
• Entity mapping: How blueprint roles (Manager, Worker) map to ServiceNow entities (Workflow, Agent)
• Tool types: Record Operation, Script, Flow, Search Retrieval, MCP Tool, etc.
• Prompting patterns: Markdown format, parameter naming, gate definitions
• Browser navigation: Verified URL patterns and Playwright selectors for AI Agent Studio
• MCP setup: servicenow-mcp-server and nowsdk-ext-mcp configuration

agent-deployment

A structured methodology for deploying AI agents on any enterprise platform:

• Phased rollout: Pilot first, validate, then expand
• Simulation-to-real progression: Prove orchestration works, then connect real data
• Test gates: Mandatory validation checkpoints before expanding scope
• Access tier adaptation: Full access, read-only, or no-access workflows
• Research-backed: Cites Microsoft, McKinsey, Forrester on why simulation-only pilots fail

research-agent

A reusable research methodology that routes queries through the right tools: