Skip to Content
View all events

Working with Agent Skills: A Hands-On Productivity Playbook

Published by O'Reilly Media, Inc.

Turn voice notes into scheduled days, meeting recordings into draft posts, and recurring prompts into a personal Skills toolkit that compounds across projects

What you’ll learn and how you can apply it

  • Decide when a recurring task should become a Skill versus a prompt
  • Convert a voice memo, meeting recording, or raw note into structured output using a chained Skills pattern with transcription and extraction stages
  • Compose the portable core together with tool-specific extras into a layered workflow
  • Leverage progressive disclosure and the SKILL.md file as a navigation map for more complex workflows
  • Start a versioned Skills library that compounds across projects

Course description

Agent Skills, which are an open, cross-agent standard now adopted by Claude Code, Codex, Cursor, GitHub Copilot, and Gemini CLI, are folders of markdown and scripts that ride on top of any modern coding agent. They have quietly become the most leveraged unit of productivity work in 2026.

In this hands-on 2-hour course with Lucas Soares, you’ll learn the practical patterns working professionals are using around Skills to compress hours of work into minutes. You’ll leave with 10 reusable productivity patterns, a clear mental model for when a recurring workflow should graduate into a Skill versus a prompt or an agent-memory file entry, and at least one starter Skill of your own that you’ll build live during the session.

This live event is for you because...

  • You’re a builder or indie developer who has built a SKILL.md or two, bounced off when it underperformed, and wants to understand what you missed.
  • You’re a product manager, researcher, or knowledge worker who already uses Claude or Cursor casually and wants to move past chat into repeatable, artifact-producing workflows that turn raw input into structured output.

Prerequisites

  • A laptop with Claude Code or an equivalent SKILL.md-compatible agent installed and a working login or API key
  • Comfort using an LLM coding agent (Claude Code, Codex, Cursor, or Gemini CLI) at least casually for real work
  • A rough understanding of the YAML frontmatter
  • Basic command-line familiarity and the ability to edit Markdown files and follow a short folder structure

Recommended preparation:

  • Confirm you can run a basic session in Claude Code (or in a SKILL.md-compatible agent such as Codex, Cursor, or Gemini CLI)
  • Clone the course repo (link will be shared a week before the session)
  • Bring one real recurring workflow from your own work that you currently solve with pasted prompts

Recommended follow-up:

Schedule

The time frames are only estimates and may vary according to how the class is progressing.

The graduation threshold and the five-primitive map (30 minutes)

  • Presentation: From prompts to Skills—the evolution of leverage; the portable-core map—Skills, Prompts, the agent-memory file, and MCP; the prompt-to-Skill graduation threshold; the five concrete graduation signals; agent-memory files versus SKILL.md; the honest counter-question
  • Demonstration: Graduating a real prompt
  • Q&A

The brain-dump to Skill pattern (30 minutes)

  • Presentation: The headline pattern—unstructured human input to structured artifact; pattern shape
  • Hands-on exercise: Record a 60-second voice note about your week and run it through a pre-built scheduling Skill
  • Demonstration: Producing a draft post, an Anki deck, and a scheduled task list from a single meeting recording
  • Q&A
  • Break

Artifact-first decomposition (35 minutes)

  • Presentation: Decomposing workflows into strategic artifacts
  • Demonstration: A working example for content generation
  • Q&A

Hoarding workflows and composing them into Skills (25 minutes)

  • Presentation: The personal hoard—capturing solved problems as composable files; keeping an “AI reference” folder; a portability cheat sheet
  • Demonstration: A tour of reference Skills catalogs across ecosystems; the Claude Code reference library, OpenAI's official Codex catalog, and a strong personal dotfiles example; turning meeting notes into tasks; creating skills through brainstorming sessions
  • Q&A

Your Instructor

  • Lucas Soares

    Lucas Soares is a machine learning engineer who has worked at K1 Digital and Biometrid, where he developed computer vision and NLP models for applications such as document verification, OCR-based applications, and recommender systems. Lucas has also developed various ML models, including neural networks, Siamese networks, convolutional neural networks, LSTMs, and genetic algorithms.

    linkedinXlinksearch

Skill covered

DevOps