- Resources
- Courses & Learning
- AI Agent Skills
Courses & LearningFree
AI Agent Skills
Turn a repeated workflow into an Agent Skill, design its discovery and activation, write SKILL.md, bundle scripts, references, and assets, then evaluate, secure, and ship it.
5 modules · 25 lessons. Readable online with no signup, or offline as a PDF or EPUB via the free newsletter.
View course (opens in new tab)Curriculum
- 1
Choose the right skill
Distinguish skills from nearby AI concepts and extract one useful workflow from real work.
- 01 Separate skills from prompts, rules, tools, and MCP
- 02 Start from a repeated real workflow
- 03 Define the outcome, inputs, and non-goals
- 04 Draw the authority boundary
- 05 Check your understanding: choose the right skill
- 2
Design discovery and activation
Structure the skill for progressive loading and precise positive and negative triggers.
- 06 Understand progressive disclosure
- 07 Choose a valid, specific name
- 08 Write the activation description
- 09 Test positive and negative triggers
- 10 Check your understanding: design discovery and activation
- 3
Write SKILL.md
Write valid metadata, an executable procedure, checkpoints, boundaries, and an output contract.
- 11 Write valid, portable frontmatter
- 12 Write a procedure, not an essay
- 13 Add checkpoints and an output contract
- 14 Teach failures, gotchas, and stopping
- 15 Check your understanding: write skill.md
- 4
Bundle scripts, references, and assets
Move deterministic work and conditional context into focused portable resources.
- 16 Decide when a script beats prose
- 17 Write a safe, self-contained script
- 18 Add focused references and assets
- 19 Control context, dependencies, and paths
- 20 Check your understanding: bundle scripts, references, and assets
- 5
Evaluate, secure, and ship
Test activation and output, review authority, package the folder, and maintain it from evidence.
- 21 Evaluate activation and output quality
- 22 Threat-model the skill
- 23 Package, document, and version the skill
- 24 Finish and maintain the release-readiness skill
- 25 Check your understanding: evaluate, secure, and ship
Author
Flavio CopesMore from Flavio Copes
AI Fundamentals
How language models work, writing prompts and project rules, curating context, agents and tools, verifying AI output, and using AI safely.
Build a Language Model on Your Mac
Train a small language model with Language Model Builder on a Mac. Covers tokens and embeddings, designing the run, pre-training and reading loss, fine-tuning with SFT and DPO, and comparing checkpoints in a lab report.
Build with MCP
Build a TypeScript MCP server with tools, resources, and prompts, move it to remote HTTP with authorization and security boundaries, then deploy and review it.
Local AI Models
Choose an open-weight model that fits your hardware, run it locally with Ollama and Node.js, build a private AI feature with structured output and fallbacks, and operate local AI responsibly.
MCP
What MCP solves, how hosts, clients, servers, tools, resources, and prompts fit together, and how to connect, configure, test, and troubleshoot an existing server you trust.
No resources match your search.