Frequently Asked Questions About Grace Leung AI Marketing Team Builder
21 answers covering everything from basics to advanced usage.
// Basics
What is the difference between system folders and working folders in the AI Marketing Team Builder?
System folders (context, SOPs, templates) contain reusable infrastructure that pre-equips agents with brand knowledge and working patterns. Working folders (ads, pages, presentations) are output destinations where agents save their produced deliverables. Mixing them breaks agent orientation — agents need to clearly distinguish between reference material they read and output locations they write to. This separation mirrors how a real business organizes its marketing department.
What is a style library in the AI Marketing Team Builder?
A style library is a folder of on-brand visual templates placed under your templates folder, accompanied by a style guide describing when each style is appropriate. It serves as visual inspiration, not an exact copy directive. Claude studies the vibe, patterns, and design language, then applies them generatively to new creative output. Using it as a pixel-perfect copy instruction produces rigid, templated work instead of creatively on-brand deliverables.
What are agent routing rules and where do I put them?
Agent routing rules are explicit instructions written into your CLAUDE.md file that tell Claude when to delegate a task to a sub-agent versus handle it with a skill call alone. The rule of thumb: synthesis tasks requiring judgment and orchestration (research, strategy) route to agents; executional, straightforward tasks (single content formats like a blog post) can be handled by skills directly. Without these rules, Claude cannot reliably self-direct on multi-step campaigns.
What does a complete AI marketing team folder structure look like?
The root project folder contains: a CLAUDE.md file, a .mcp.json file (optional), a marketing subfolder with system folders (context/, SOPs/, templates/) and working folders (ads/, pages/, presentations/), a Skills/ folder created by Claude for custom skill assets, and an Agents/ folder containing one markdown file per agent defining its role, skills, and responsibilities. Templates can include subfolders like social-creatives/ with style guides. Brand context files go in context/ before anything else is built.
Is the AI Marketing Team Builder only for large marketing teams?
No — it's particularly powerful for solo marketers and small teams who need to operate across multiple marketing functions without hiring specialists. A solo marketer can build five AI agents that collectively cover research, strategy, content, design, and analytics. The system acts as force multiplication, turning one person into a functional marketing department. The key requirement is not team size but having repeatable marketing workflows worth systematizing.
How long does it take to build the full AI marketing team?
Initial setup — folder structure, context loading, CLAUDE.md creation, and installing official skills — takes about 30-60 minutes. Building custom skills using the Reference-Based Method takes 15-30 minutes per skill. Creating and configuring agents takes about 10-15 minutes each. A complete team with five agents and 8-10 custom skills can be operational in a single focused afternoon. Once built, a full campaign package runs in roughly 10 minutes.
// How To
How do I set up MCP connections for external tools in my AI marketing team?
Create a .mcp.json file in your project root folder. This file declares which MCP servers Claude can connect to, along with the required API keys for services like image generation. After creating the file, restart Claude Code and verify your connections work by running the /mcp slash command. Only then should you build skills that depend on those external tools. Skipping verification will cause skill failures at runtime.
How do I install official skills in Claude Code for marketing?
Use the /plugin slash command to add Anthropic's official skills GitHub repository marketplace. Search for and install the document skills pack, which includes foundational skills for common document types. Confirm installation by checking that all official skills appear under the slash command menu. These official skills serve as base templates you can extend into custom branded skills using the Reference-Based Method.
How do I connect a Notion task board to my AI marketing team?
Set up a Kanban board in Notion with columns for To-Do and Complete, and fields for task title, details, and priority. Then prompt Claude to scan pending tasks, assign the appropriate agent based on task type, execute in priority order, and update each task's status to Complete with output file paths included. This creates a bridge where human teammates drop tasks and AI agents execute them, operating as a 24/7 execution layer alongside your team.
How many agents should I create for my AI marketing team?
The recommended structure is five agents: Data Analyst (campaign reporting, data visualization), Content Creator (blog writing, keyword research, lead magnets), Market Researcher (audience and trend research), Creative Designer (social creatives, branded decks), and Campaign Strategist (campaign briefs, landing pages). The key constraint is non-overlapping roles — each agent should have a coherent cognitive focus. You can start with fewer and add agents as your workflow demands grow.
// Troubleshooting
Why is my AI marketing team producing generic, off-brand content?
The most common cause is skipping the context loading step. Agents built without brand voice guides, style guides, and marketing strategy documents produce fundamentally generic output. Load all brand context files into the dedicated context folder before building any skills or agents. This pre-equips the entire team with brand knowledge. Also verify your style library is loaded for creative skills and that your Reference-Based Method analysis reports are thorough.
Why does Claude lose focus when I give it complex marketing tasks?
This happens when too many skills are piled into one agent or one conversation without proper role separation. Claude loses focus the same way one person cannot be writer, analyst, and designer simultaneously. The fix is to keep agent roles non-overlapping with coherent skill sets, and to add explicit routing rules in CLAUDE.md so Claude delegates synthesis tasks to the right sub-agent instead of trying to handle everything in a single thread.
What do I do when my Claude Code remote session fills up its context?
Type 'clear conversation' to reset the context window rather than continuing in an overloaded state. Continuing with a full context causes degraded output quality and unreliable agent routing. After clearing, you can resume dispatching tasks normally. All previous work is saved locally in your project folders, so you don't lose any produced deliverables. Disconnect via the three-dot menu and archive when your session is complete.
My skills produce output that doesn't match my brand templates — what went wrong?
You likely built the skill from a blank prompt instead of using the Reference-Based Method. To fix this: place the branded template in your templates folder, prompt Claude to analyze it and produce a detailed analysis report capturing layout rules, color usage, font hierarchy, and structural patterns, then rebuild the skill from the analysis. This anchors quality to real brand standards and typically achieves 90%+ alignment on first run.
// Comparisons
How does the AI Marketing Team Builder compare to using separate AI tools for each marketing task?
Separate tools (one for writing, one for design, one for research) lack shared context, brand memory, and coordinated output. The AI Marketing Team Builder creates a unified system where all agents share the same brand context, follow consistent style standards, and produce thematically connected deliverables. A campaign brief written by the strategist agent directly informs the content creator and designer agents, producing cohesive campaigns rather than disconnected assets.
How is the AI Marketing Team Builder different from using CrewAI or AutoGen for marketing agents?
CrewAI and AutoGen are code-first frameworks requiring Python programming to define agents, tasks, and orchestration logic. The Grace Leung AI Marketing Team Builder operates entirely within Claude Code's native features — /agents, /plugin, CLAUDE.md, and the file system — requiring no external coding. It's designed for marketers, not developers. The tradeoff is less programmatic control but dramatically faster setup and a system that non-technical team members can operate and extend.
// Advanced
Can I add new skills and agents to my AI marketing team after the initial build?
Yes — the system is explicitly designed for iterative expansion. Add new skills by repeating the Reference-Based Method for any new workflow. Create new agents via the /agents command when a new role is needed. The critical step most people miss: update CLAUDE.md every time you add a skill or agent. CLAUDE.md is a living document, not a one-time configuration. Without updates, Claude won't know the new skill or agent exists and can't route tasks to it.
How do I decide whether a task should go to an agent or be handled by a skill alone?
Use the synthesis-versus-execution rule encoded in your CLAUDE.md routing rules. Tasks requiring synthesis, judgment, or orchestration across multiple skills — like market research, campaign strategy, or multi-format content planning — should route to an agent. Straightforward executional tasks — like generating a single blog post, creating one social graphic, or formatting a document — can be handled by a skill call alone. This keeps the system efficient without over-engineering simple tasks.
What is the Map → Skill → Agent → Team sequence and why can't I skip steps?
This is the four-layer build sequence: first map every repeatable marketing task, then convert each into a reusable skill, then group skills into focused agent roles, then connect agents as a coordinated team. Skipping layers — for example, jumping straight to creating agents without defining skills first — produces unfocused agents with unclear responsibilities. The sequence ensures each layer builds on a solid foundation, resulting in a reliable, self-directing system.
How do I write effective CLAUDE.md routing rules for my marketing team?
Structure your routing rules as conditional logic: 'When a task involves [task type], delegate to [Agent Name]. When a task only requires [specific output], call [Skill Name] directly.' Include fallback rules for ambiguous tasks. Specify which agent serves as the coordinator for multi-deliverable campaigns. Be explicit about handoff patterns — for example, 'Market Researcher completes research before Campaign Strategist begins brief.' Test routing with a full campaign brief and refine based on observed behavior.
Can the AI Marketing Team Builder handle multiple brands or clients?
Yes, by creating separate project folders for each brand or client, each with its own context folder, templates, style library, CLAUDE.md, and agent configurations. Each project folder operates as an isolated system with brand-specific knowledge. You cannot share agents across project folders natively, but you can replicate your agent structure and customize the context and templates per brand. This mirrors how a real marketing agency organizes client work.