Creating Agents with BMAD

🎥 Video lesson — video not yet migrated (source Skool video id: 100b342ccdaf475e8027c2f5d24974eb).

Creating Your First BMAD Agent — Building the Content Strategist

This video walks through the full process of creating a custom agent inside the BMAD framework. We build out the Content Strategist agent (codename: Scout) from the module we created in the previous video, covering everything from discovery through to a working slash command.

Where we're at: In the last video we built the Content Creation Suite module, which gave us 6 agent placeholders and 22 workflow specs. Now we're going one level deeper — actually building out the first agent so it's functional with commands, memory, and brand context.

Step-by-step process:

  1. Run /bmb-create-agent (or use the Agent Builder) to kick off the agent creation workflow
  2. Load the existing agent spec using the @ symbol to pass the file into the conversation
  3. Go through the discovery process — define the target user (solo creator), platforms (YouTube, LinkedIn, X, Instagram, TikTok, email, blogs), and skill level (medium)
  4. Review the generated agent metadata, commands, and sidecar documents
  5. Move the completed agent from bmad-output/ into the actual module's agents directory
  6. Manually set up the slash command in .claude/commands/ so you can trigger the agent with /content-strategist
  7. Populate the sidecar files (brand guidelines, ICP profile) with real business context

Key concepts covered:

  • Sidecar documents — separate files linked to an agent that store brand guidelines, ICP profiles, and content history. These act as persistent memory across conversations and projects
  • Activation loading — every time you invoke the agent, it automatically loads brand guidelines and ICP context before you even give it a command, so everything stays on-brand
  • Context injection — each command triggers a specific workflow file, keeping conversations lean by only loading what's needed for that particular task
  • BMAD Output folder — a staging area where the framework generates and brainstorms documents before moving finalised versions into the proper project structure

Commands built into the Content Strategist agent:

  • Chat with the agent
  • Competitive research (triggers workflow)
  • Content ideation (triggers workflow)
  • Content repurpose (triggers workflow)
  • Workflow status check

Tips from the video:

  • Use Claude Desktop for quick tasks — generating brand guidelines, drafting emails, or summarising documents is faster in Claude Desktop. Save Claude Code for building workflows and writing code
  • Nothing is deterministic — sometimes a step in the workflow gets missed (like auto-creating the slash command). Understanding what's happening under the hood means you can fix things when they go off track
  • Populate your sidecar files properly — the more context you give your agent about your brand and ICP upfront, the better every output will be. Scrape your own website, pull from existing docs, upload YouTube analytics
  • You can run two Claude Code conversations side by side in Cursor by dragging terminals next to each other — handy for building in one while testing in another

What still needs to be done: The agent is built and triggerable via slash command, but the individual workflows (competitive research, content ideation, content repurpose, workflow status) still need to be created. That's covered in the next video.

Next up: Building the workflows that power the Content Strategist agent's commands.

← Previous
Next →