Every agent and sub-agent in this collection is designed to be used out of the box — but the real power comes from adapting them to your specific workflow, voice, and context. This guide walks you through the most common customization scenarios.
Change only what you need. The agents in this collection have been written and tested to work well as-is. The more you change, the more you should test. Start with small adjustments and verify the agent's behavior before making bigger changes.
These changes are safe, low-risk, and almost always worth doing.
comms_channels: [telegram, discord]
Update this to match the channels you actually use. If you only use Telegram, remove the others to keep things clean.
memory_profile: "deep"
Options:
minimal— Use when privacy matters or you want a fresh experience each sessionstandard— Good default for most personal agentsdeep— Best for professional agents that need to remember contacts and commitments
sub_agents: [email-manager, web-researcher]
Only list sub-agents you have installed. Listing a sub-agent that isn't installed will cause delegation failures.
model_recommendations:
primary: "gpt-4o"
fallback: "claude-3-haiku"
Use the model that gives the best results for your use case. Code review agents often benefit from larger context models. Briefing agents can use faster, smaller models.
The ## Persona section defines who the agent is. You can adapt it to match your style.
Simply change the name in both the frontmatter and the Persona section:
name: "Alex"
## Persona
You are Alex, a warm and efficient personal assistant...
Find the tone description in ## Communication Style and modify it:
Before:
Warm, concise, and practical.
After:
Formal and efficient. Use professional language at all times.
Avoid casual phrasing. Address the user as "you" not by first name.
In the Persona section, add context about your industry or role:
You are aware that the user works in the healthcare industry
and is familiar with HIPAA compliance requirements.
When handling patient-adjacent topics, always flag
privacy considerations.
Simply add a new bullet to ## Core Behaviors:
- Always respond in Spanish unless the user writes in another language
- Never suggest scheduling meetings before 9:00 AM or after 6:00 PM
- When the user says "EOM" at the end of a message, that means no reply is needed
Delete the bullet that describes the behavior you want to remove. Be careful not to remove behaviors that are referenced in other sections.
For heartbeat-enabled agents, find the time reference in ## Heartbeat Behavior and update it:
Aria's heartbeat fires every morning at 6:30 AM local time (changed from 7:00 AM default).
Then update your OpenClaw heartbeat configuration to match.
If you want to extend what an agent can do:
- Add the new capability to
## Capabilities - If it requires a new skill, add the skill name to
skills_requiredin the frontmatter - If appropriate, add a new
## Core Behaviorsbullet describing how the agent should use this capability - Add an example interaction demonstrating the new capability
When deploying agents for a whole team (especially Velocity for scrum teams), consider:
On first use, give the agent context about your team:
"Velocity, our team is: Alice (frontend lead), Bob (backend), Carol (QA). We run 2-week sprints. Our current sprint started March 25 and ends April 7. Our standup is at 9:30 AM Pacific in #team-standup."
Tell the agent your preferences so it remembers them:
"Maxwell, my working hours are 8 AM to 5 PM Pacific. Don't schedule meetings before 8 AM or after 4 PM. I prefer meetings clustered in the morning."
Paste your brand guidelines and tell Lyra to remember them:
"Lyra, please remember this as our brand voice guide: [paste guidelines]. Use this for all content you write for us."
If you make significant changes to an agent, bump the version number in the frontmatter:
version: "1.1.0" # was 1.0.0
This helps you track what changed and makes it easier to see which agents have been customized vs. running the original.
If you create a meaningfully improved or specialized version of an agent, consider contributing it back to the community as a new agent. See CONTRIBUTING.md for guidelines. A well-crafted "healthcare executive assistant" or "game developer code reviewer" could be valuable to others!