Building Agent Skills
Design patterns, code examples, and proven solutions for building AI agent skills. Step-by-step guides for developers.
Building agent skills with MCP
The Model Context Protocol is how tools talk to AI models. Here's what it is, why it matters, and how to build your first MCP server.
Observability for agents
Your agent did something weird and you have no idea why. Logging, tracing, and debugging patterns that make agent behavior understandable.
Building a Slack bot with agent skills
A step-by-step walkthrough of building a Slack bot that can search your docs, answer questions, and take actions, using agent skills and the Slack API.
Building a data pipeline tool
An ETL-style data pipeline tool with validation stages and human-in-the-loop confirmation, demonstrating multi-step workflows and safety patterns in Python.
AI agents for DevOps and SRE teams
Incident response, runbook automation, log analysis, and alert triage. How agent skills fit into the on-call workflow without making things worse.
Skill composition
Orchestrating multiple skills together: dependency resolution, parallel execution, shared context, and building complex capabilities from simple parts.
Building a code review tool
A multi-step code review tool that diffs changes, analyzes code quality, and generates structured feedback, demonstrating skill composition and error handling.
Context management for AI agents
Strategies for working within context window limits: summarization, selective loading, and memory patterns for agent skills.
Multi-step workflows for AI agents
Patterns for chaining skills into complex, multi-step workflows with state management, branching logic, and recovery strategies.
Human-in-the-loop patterns for AI agents
Patterns for involving humans in agent workflows: approval gates, progressive autonomy, and knowing when to escalate.
AI agent skill anti-patterns to avoid
Common mistakes in agent skill design: the god tool, leaky abstractions, over-parameterization, and patterns that lead to unreliable agents.
Error handling patterns for AI agents
How to build agent skills that handle failures gracefully: retry strategies, fallbacks, partial completion, and informative error responses.
Building a file search tool
A complete, annotated implementation of a file search tool with glob pattern matching, result formatting, and thorough tests.
How skills use tools
How to write skills that call the right tools with the right inputs and handle what comes back. Skills encode judgment; tools execute.