# How to Customize Agent Personalities and Workflows for Specific Project Needs

> Customize agent personalities and workflows for your project needs by editing Markdown files in Agency-Agents. Modify YAML for identity and update workflow phases without code changes.

- Repository: [Michael Sitarzewski/agency-agents](https://github.com/msitarzewski/agency-agents)
- Tags: how-to-guide
- Published: 2026-03-09

---

**You customize agent personalities and workflows in the Agency-Agents framework by editing Markdown definition files—modifying YAML frontmatter for identity traits and updating workflow phases in the orchestrator—without requiring any code changes or redeployment.**

The Agency-Agents repository by msitarzewski provides a modular system where every specialist agent is defined as a self-contained Markdown file. To customize agent personalities and workflows for specific project needs, you edit these declarative configuration files that the orchestration layer reads at runtime, enabling instant adaptation to unique branding, technical, or procedural requirements.

## Understanding the Agent File Structure

Each agent resides in a Markdown file with YAML frontmatter defining its identity, followed by sections for personality, mission, and workflows. The frontmatter schema includes:

```yaml
---
name: <Agent-Name>
description: <Brief role summary>
color: <UI accent>
---

```

Below the frontmatter, the file defines **personality** traits, **core mission**, **critical rules**, and often a **workflow template** that the orchestrator executes.

## Customizing Agent Personalities

### Modifying Frontmatter Attributes

Edit the YAML frontmatter in the agent's Markdown file to change basic identity attributes. For example, in [`design/design-whimsy-injector.md`](https://github.com/msitarzewski/agency-agents/blob/main/design/design-whimsy-injector.md), you can update the `color` field from pink to teal to match brand guidelines, or modify the `description` to reflect a different creative focus.

### Updating Personality Traits and Voice

The personality section typically appears as a bullet list defining behavioral traits. In [`design/design-whimsy-injector.md`](https://github.com/msitarzewski/agency-agents/blob/main/design/design-whimsy-injector.md), lines 12-15 define the default personality as "Playful, creative, strategic". Replace these traits with project-specific attributes like "Bold, witty, inclusive" to shift the agent's communication style and decision-making approach.

### Adjusting Micro-Copy and Design Tokens

For agents that generate UI text, modify the micro-copy library section (starting around line 95 in the Whimsy Injector file) to embed new voice and tone guidelines. This ensures consistent language across all generated deliverables, from button labels to error messages.

## Tailoring Workflows for Project Requirements

### Modifying Workflow Phases

The **Agents Orchestrator** in [`specialized/agents-orchestrator.md`](https://github.com/msitarzewski/agency-agents/blob/main/specialized/agents-orchestrator.md) defines project workflows through sequential phases. Locate the `## 🔄 Your Workflow Phases` section (lines 51-94) to add, remove, or reorder phases. For example, insert a "User Research & Insights" phase after Phase 1 by adding a new `### Phase 2:` block with bash-style commands to spawn research agents.

### Changing Agent Assignments

Within each phase block, modify the quoted spawn commands to assign different specialist agents. Replace `"Please spawn a **marketing-growth-hacker**..."` with `"Please spawn a **ux-researcher**..."` to redirect workflow tasks to appropriate specialists based on project needs.

### Configuring Quality Gates and Decision Logic

Adjust the `### Critical Rules` section (lines 37-45) to tighten or loosen retry limits, evidence requirements, or escalation triggers. Modify the markdown tables under `### Decision Logic` (lines 108-145) to establish new pass/fail criteria or additional automation steps, ensuring quality gates align with specific project standards.

## Activating Customized Agents

After modifying agent definitions, use activation prompts stored in [`strategy/coordination/agent-activation-prompts.md`](https://github.com/msitarzewski/agency-agents/blob/main/strategy/coordination/agent-activation-prompts.md) to invoke the customized behavior. Create or edit prompts to include project-specific context such as target audience, tone requirements, or KPIs. For example:

```markdown

# activation-prompt-marketing-growth-hacker.md

You are a **Growth Hacker** tasked with acquiring 1,000 qualified leads in 30 days for a B2B SaaS product. Use the following KPIs:
- CAC ≤ $50
- Conversion Rate ≥ 5 %
- Channels: LinkedIn, Content Marketing, Paid Search

```

Launch the agent by referencing the updated prompt:

```bash
Please spawn a marketing-growth-hacker agent with the prompt "Execute the above growth plan for the upcoming product launch."

```

## Summary

- **Agent personalities** are customized by editing Markdown frontmatter (name, description, color) and personality trait lists in individual agent files like [`design/design-whimsy-injector.md`](https://github.com/msitarzewski/agency-agents/blob/main/design/design-whimsy-injector.md).
- **Workflows** are tailored by modifying phase definitions, agent assignments, and quality gates in [`specialized/agents-orchestrator.md`](https://github.com/msitarzewski/agency-agents/blob/main/specialized/agents-orchestrator.md).
- **Activation** occurs through project-specific prompts in [`strategy/coordination/agent-activation-prompts.md`](https://github.com/msitarzewski/agency-agents/blob/main/strategy/coordination/agent-activation-prompts.md) that inject context at runtime.
- All changes are declarative and take effect immediately without code changes or redeployment.

## Frequently Asked Questions

### Do I need to restart the system after editing agent files?

No. The Agency-Agents framework reads Markdown definition files at runtime when agents are spawned. Changes to personality traits in [`design/design-whimsy-injector.md`](https://github.com/msitarzewski/agency-agents/blob/main/design/design-whimsy-injector.md) or workflow phases in [`specialized/agents-orchestrator.md`](https://github.com/msitarzewski/agency-agents/blob/main/specialized/agents-orchestrator.md) take effect immediately on the next activation without requiring system restarts or redeployment.

### Can I create entirely new agent types for specialized roles?

Yes. Create a new Markdown file following the established schema with YAML frontmatter (name, description, color) followed by sections for personality, mission, and workflow rules. Place the file in the appropriate directory (e.g., `design/` for creative agents or `engineering/` for technical agents). The orchestrator in [`specialized/agents-orchestrator.md`](https://github.com/msitarzewski/agency-agents/blob/main/specialized/agents-orchestrator.md) can then reference your new agent in workflow phases.

### How do I ensure quality standards are maintained when customizing workflows?

Modify the `### Critical Rules` section in [`specialized/agents-orchestrator.md`](https://github.com/msitarzewski/agency-agents/blob/main/specialized/agents-orchestrator.md) (lines 37-45) to define retry limits, evidence requirements, and escalation triggers. Additionally, update the `### Decision Logic` tables (lines 108-145) to establish clear pass/fail criteria for each phase. These quality gates are evaluated at runtime to ensure customized workflows maintain project standards.

### Where should I store project-specific activation prompts?

Store custom activation prompts in [`strategy/coordination/agent-activation-prompts.md`](https://github.com/msitarzewski/agency-agents/blob/main/strategy/coordination/agent-activation-prompts.md) or create separate Markdown files in the `strategy/coordination/` directory. Each prompt should reference the target agent name and include project-specific context such as target audience, tone requirements, KPIs, or constraints. The orchestrator reads these prompts when processing spawn commands.