Your AI coding agent training is complete, but your learning journey is just beginning. This page collects resources and paths to help you continue growing your AI-assisted development skills.
Your Tool’s Official Documentation
The single most important resource is your agent’s own documentation. Bookmark it and check it regularly — these tools move fast and new capabilities appear frequently.
What to look for in the docs:
- Configuration: How to write effective context files (like AGENTS.md) and set permissions
- Skills / slash commands: Creating and sharing reusable workflows
- MCP Servers: Connecting the agent to external tools and services
- Hooks: Event-driven automation (pre/post tool use, session events)
- Subagents: Custom agent definitions and parallel work
- Best Practices: Tips for effective use from the people who built the tool
- Changelog / release notes: New features ship regularly; stay current
Check the release notes at least monthly. Capabilities that did not exist when you took this training may now solve problems you have been working around.
The AGENTS.md Open Standard
AGENTS.md is the vendor-neutral open standard for giving coding agents persistent project context. The spec covers file hierarchy (global, project, local), what to include, and how to evolve the file over time. If you move between tools or use multiple agents, a well-written AGENTS.md works across all of them.
Model Context Protocol (MCP)
The official MCP specification and documentation. Useful when building custom MCP servers or evaluating third-party servers for your team.
MCP Server Registry
github.com/modelcontextprotocol/servers
Community-maintained registry of MCP servers. Browse available servers for databases, APIs, and services. Check here before building a custom server — someone may have already built what you need.
Community and Practice
Practice on your own codebase
The fastest way to improve is daily use on real work. After training, the most productive engineers:
- Use the agent for every task they would have done manually
- Track which prompts work well and add them to their team’s context file
- Build a personal library of prompts that reliably produce good output for their stack
- Share effective patterns with their team
Community resources
Search for communities around agentic coding on:
- GitHub Discussions on the MCP server registry and related repos
- Reddit communities for your primary language or framework (most now have active AI-assisted development threads)
- Discord servers for your tech stack — AI coding workflows come up frequently
- Conference talks on AI-assisted development (major tech conferences now include dedicated tracks)
Advanced Topics to Explore
Once you are comfortable with daily agent use, consider exploring these areas:
Custom MCP Servers
Build MCP servers that connect your coding agent to your internal systems — databases, monitoring tools, deployment pipelines, and internal APIs. The MCP TypeScript SDK makes it straightforward to expose any data source as an agent tool.
- MCP TypeScript SDK: github.com/modelcontextprotocol/typescript-sdk
Custom Skills
Create reusable skill definitions that encode your team’s best practices into repeatable workflows. Skills live in .agent/skills/ (or the equivalent directory for your tool) and can be invoked with slash commands.
CI/CD Integration
Integrate your coding agent into continuous integration pipelines for automated code review, PR description generation, and test generation on changed files. Use non-interactive (-p) mode with --output-format json for machine-readable output.
Multi-Agent Workflows
Explore patterns where agents run in parallel — one investigating a bug while another writes tests, or multiple agents exploring different parts of a large codebase simultaneously. Your agent’s subagent or parallel execution documentation covers the specifics.
Recommended Learning Path
First Month After Training
Week 1: Focus on daily use. Refer back to the cheat sheets and before and after examples when stuck.
Week 2: Read your tool’s full documentation end-to-end. You will find at least three features you did not know about.
Week 3: Build something with MCP. Connect the agent to one internal system — a database, a monitoring tool, or an internal API.
Week 4: Review your productivity metrics. Share results and insights with your broader engineering organization.
Month Two and Beyond
- Build a custom MCP server for one of your internal systems
- Contribute to and maintain the team AGENTS.md
- Mentor other engineers on AI-assisted development
- Create custom skills for your team’s most common workflows
- Host a lunch-and-learn to share your experience with other teams