Open-source AI agent skills for enterprise deployment. Built for coding agents (Claude Code, Codex, Cursor, Copilot, Windsurf) using the Agent Skills open standard.
7 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 |
Pre-designed agent team architectures for common enterprise use cases. Each pattern describes the full agent team, roles, orchestration, and integration points. Use as a starting point, or generate a full customized blueprint via the Agent Blueprint MCP server.
| Pattern | What it does |
|---|---|
| it-service-desk-triage | 5-agent IT help desk: triage, classification, knowledge search, auto-resolution, escalation |
| procurement-rfx-processing | 6-agent RFP/RFQ/RFI lifecycle: intake, requirements, vendor matching, compliance, evaluation, communication |
| customer-onboarding | 5-agent onboarding pipeline: intake, verification, provisioning, communication, handoff |
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.
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/Clone this repo and point your agent to the SKILL.md files:
git clone https://github.com/agent-blueprint/agent-blueprint-skills.git .agent-skillsThe SKILL.md format is an open standard. Any agent that reads markdown can use these skills.
Skills are in plugins/agent-blueprint-skills/skills/.
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.mdA 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)
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
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
A reusable research methodology that routes queries through the right tools:
- Depth routing: WebSearch (free) to Brave to Perplexity to Gemini (expensive)
- Escalation rules: Start cheap, escalate when results are thin
- Supplemental sources: Financial filings, careers pages, reviews, competitive intel
- Caching: Avoid redundant queries with freshness-aware cache
- Tool-agnostic: Works with whatever MCP servers you have configured
A 5-agent architecture for automating IT service desk operations:
- Triage Agent: Extracts structured data, sets priority, routes tickets
- Classification Agent: Deep categorization, duplicate detection, assignment group recommendation
- Knowledge Search Agent: Semantic search across knowledge base, relevance ranking
- Resolution Agent: Auto-resolves known patterns (password resets, access requests, known errors)
- Escalation Agent: Packages context for human analysts, monitors SLA compliance
- Orchestration: Routing pattern with Knowledge Search as utility agent
A 6-agent architecture for end-to-end RFx lifecycle automation:
- Intake Agent: Extracts requirements from source documents, creates structured RFx record
- Requirements Agent: Validates requirements, identifies gaps and conflicts, builds scoring rubric
- Vendor Matching Agent: Identifies qualified vendors from database, past performance, compliance status
- Compliance Agent: Validates regulatory and policy compliance at every stage
- Evaluation Agent: Scores vendor responses, produces comparative analysis and award recommendation
- Communication Agent: Manages all stakeholder and vendor communications
- Orchestration: Orchestrator + Workers with parallel vendor matching and compliance
A 5-agent architecture for customer onboarding pipeline automation:
- Intake Agent: Receives applications from any channel, validates completeness, sets risk tier
- Verification Agent: Identity, compliance, KYC/AML, document validation
- Provisioning Agent: Account creation, billing, access, integrations (parallel tasks)
- Communication Agent: Status updates, document requests, activation packages
- Handoff Agent: Transitions to ongoing service with structured CSM package
- Orchestration: Pipeline with parallel branches, async waits for customer actions
These skills follow the Agent Skills open standard, supported by:
| Platform | Skill location | How it works |
|---|---|---|
| Claude Code | .claude/skills/ or plugin |
Loaded via plugin system or project skills |
| OpenAI Codex | .agents/skills/ |
Same SKILL.md format, automatic discovery |
| Cursor | Project context | Read SKILL.md as context |
| GitHub Copilot | .github/skills/ |
Agent Skills standard |
| Any agent | Direct file read | SKILL.md is just markdown |
These community skills cover the fundamentals. Agent Blueprint provides:
- Full AI advisory pipeline: Business profile analysis, AI readiness assessment, use case generation, agent blueprints, business cases, and implementation plans
- Expert-level deployment skills: Battle-tested patterns, version-specific configuration, security hardening, and proprietary gotchas from real enterprise deployments
- Living blueprints: Ongoing advisory with bidirectional state sync, performance monitoring, and strategic recommendations
- MCP server + CLI:
agentblueprintnpm package for programmatic blueprint access
Contributions welcome. Found a missing URL pattern, a better prompting technique, or a useful deployment tip? Open a PR.
Please keep contributions:
- Vendor-agnostic in deployment and research skills
- Verified against real instances (for ServiceNow patterns)
- Concise -- these skills are loaded into LLM context windows, every token counts
Apache 2.0. See LICENSE.
Built by Agent Blueprint (NowGentic LLC).