Skip to content
ClawHalla

Customization

ClawHalla is designed to be deeply customizable. Everything from agent personalities to project methodologies is defined in editable files — no code changes required for most customizations.

Customizing Agent Personas

Agent personas are defined by markdown files in their workspace directory. Edit these files to change how an agent thinks, communicates, and operates.

SOUL.md — Personality and values

This is the agent’s core identity. Changes here affect tone, decision-making, and communication style.

SOUL.md
# SOUL.md


## Core Truths
- Ship fast, iterate faster. Perfect is the enemy of done.
- Code reviews are conversations, not gatekeeping.
- Every PR should make the codebase better than it was.


## Communication Style
- Direct and concise. No filler.
- Use code examples over lengthy explanations.
- Disagree constructively -- explain why, not just what.


## Boundaries
- Never push to main without CI passing
- Ask before refactoring code you did not write
- Private conversations stay private

AGENTS.md — Operating instructions

This controls what the agent does and how it handles different situations:

AGENTS.md
## Squad Delegation
| Domain | Action |
|--------|--------|
| Bug reports | Investigate, fix, write tests |
| Feature requests | Create epic, break into stories |
| Code reviews | Review for correctness, style, security |
| Deployments | Prepare, test, wait for approval |


## Context Limits
- Alert at 40% context usage
- Hard stop at 75%
- Max 3 retries on any failed task

IDENTITY.md — Name and metadata

Change the agent’s name, emoji, and model assignment:

IDENTITY.md
- **Name:** Valkyrie
- **Role:** Senior Backend Developer
- **Emoji:** ⚔️
- **Model:** `anthropic/claude-sonnet-4-6`

Creating Squad Packs

A squad pack is a pre-configured set of agents that work together on a domain. Packs are stored as YAML files in ~/.openclaw/workspace/packs/.

Pack structure

packs/marketing-squad.yaml
name: Marketing Squad
version: "1.0.0"
description: Content marketing team with SEO, social media, and analytics
author: your-name


squad:
  id: marketing_squad
  name: Marketing Squad
  domain: Content marketing, SEO, social media
  chief: content-chief


agents:
  - id: content-chief
    name: "Content Chief"
    emoji: "📝"
    role: content_director
    model: claude-sonnet-4-6
    tier: 2
    reportsTo: claw
    skills:
      - content-strategy
      - editorial-calendar
    description: "Manages the content pipeline and editorial strategy"


  - id: seo-writer
    name: "SEO Writer"
    emoji: "🔍"
    role: seo_specialist
    model: claude-sonnet-4-6
    tier: 3
    reportsTo: content-chief
    skills:
      - seo-optimization
      - keyword-research
    description: "Writes SEO-optimized content for blog and landing pages"


  - id: social-manager
    name: "Social Manager"
    emoji: "📱"
    role: social_media_manager
    model: claude-haiku-4-5
    tier: 3
    reportsTo: content-chief
    skills:
      - social-scheduling
      - engagement-tracking
    description: "Manages social media posts and engagement"

Installing a pack

Via Mission Control:

  1. Go to Marketplace in the sidebar
  2. Browse or upload a pack
  3. Click Install — agents are created automatically

Via API:

Terminal window
curl -X POST http://localhost:3000/api/packs \
  -H "Content-Type: application/json" \
  -d @packs/marketing-squad.yaml

Adding Skills

Skills are markdown files that provide domain-specific knowledge to agents. They live in the skills/ directory, organized by category.

Workspace-level skills (shared)

~/.openclaw/workspace/skills/
├── engineering/
│   ├── SKILL.md
│   ├── docker-best-practices.md
│   └── ci-cd-patterns.md
├── content/
│   ├── SKILL.md
│   └── seo-checklist.md
└── research/
    ├── SKILL.md
    └── link-transcription.md

Agent-level skills (specific)

~/.openclaw/workspace/squads/dev/backend/
└── skills/
    ├── postgres-optimization.md
    └── api-design-patterns.md

Agent-level skills take precedence over workspace-level skills.

Writing a skill

Each skill category should have a SKILL.md that describes when and how to use the skills:

skills/engineering/SKILL.md
# Engineering Skills


## When to use
- Writing or reviewing code
- Setting up infrastructure
- Debugging production issues


## Available knowledge
- docker-best-practices.md -- Container patterns and anti-patterns
- ci-cd-patterns.md -- Pipeline design for GitHub Actions

Individual skill files contain the actual knowledge:

skills/engineering/docker-best-practices.md
# Docker Best Practices


## Multi-stage builds
Always use multi-stage builds to reduce image size:


\`\`\`dockerfile
FROM node:24-alpine AS builder
WORKDIR /app
COPY package*.json ./
RUN npm ci --production
COPY . .
RUN npm run build


FROM node:24-alpine
COPY --from=builder /app/dist ./dist
COPY --from=builder /app/node_modules ./node_modules
CMD ["node", "dist/index.js"]
\`\`\`


## Health checks
Always include a HEALTHCHECK instruction...

Importing skills via API

Terminal window
curl -X POST http://localhost:3000/api/skills/import \
  -H "Content-Type: application/json" \
  -d '{
    "name": "react-patterns",
    "category": "engineering",
    "content": "# React Patterns\n\n## Component design..."
  }'

Custom Boards and Methodology

ClawHalla uses AI-AGIL, an agile methodology adapted for AI agent teams. Customize it by editing the files in ~/.openclaw/workspace/methodology/.

Board structure

Boards follow a kanban flow with YAML files:

boards/
├── backlog/
│   └── tasks.yaml
├── doing/
│   └── tasks.yaml
├── review/
│   └── tasks.yaml
└── done/
    └── tasks.yaml

Custom task templates

Edit ~/.openclaw/workspace/skills/ai-agil/templates/:

templates/task-template.md
# Task: {TASK_ID}


## Description
{description}


## Acceptance Criteria
- [ ] {criterion_1}
- [ ] {criterion_2}


## Assigned To
{agent_id}


## Priority
{priority}


## Definition of Done
- [ ] Code written and tested
- [ ] PR created and reviewed
- [ ] Documentation updated
- [ ] Approved by chief

Sprint configuration

templates/sprint-template.md
# Sprint: {SPRINT_NAME}


## Duration
{start_date} to {end_date}


## Goals
1. {goal_1}
2. {goal_2}


## Stories
{story_list}


## Capacity
{agent_capacity_table}

Theming Mission Control

Mission Control is built with Next.js and Tailwind CSS. Customize its appearance by modifying the source in apps/mission-control/.

Colors and styles

Edit the global CSS:

apps/mission-control/src/app/globals.css
:root {
  --primary: #c9a959;      /* Gold accent */
  --background: #0a0a0a;   /* Dark background */
  --card: #1a1a1a;          /* Card background */
  --text: #e5e5e5;          /* Primary text */
  --muted: #888888;         /* Muted text */
}

Dashboard layout

Edit the dashboard page at apps/mission-control/src/app/(dashboard)/dashboard/page.tsx to add or remove widgets.

Rebuilding after changes

Terminal window
cd apps/mission-control
pnpm build
pnpm start

Adding API Endpoints

Extend Mission Control with custom API routes. Create new files in apps/mission-control/src/app/api/:

apps/mission-control/src/app/api/custom/route.ts
import { NextResponse } from 'next/server';


export async function GET() {
  return NextResponse.json({
    ok: true,
    message: 'Custom endpoint working',
  });
}


export async function POST(req: Request) {
  const body = await req.json();
  // Your custom logic here
  return NextResponse.json({ ok: true, data: body });
}

Access it at http://localhost:3000/api/custom.

Creating Cron Jobs and Heartbeats

Cron jobs

Cron jobs run agent tasks on a schedule. Create them via CLI or API:

Terminal window
# Daily standup report at 9 AM
openclaw cron add \
  --name "daily-standup" \
  --agent odin \
  --cron "0 9 * * *" \
  --message "Generate a standup report of all squad activity in the last 24h" \
  --session isolated \
  --model sonnet

Via Mission Control API:

Terminal window
curl -X POST http://localhost:3000/api/crons \
  -H "Content-Type: application/json" \
  -d '{
    "name": "daily-standup",
    "agentId": "odin",
    "cron": "0 9 * * *",
    "message": "Generate standup report for dev squad",
    "model": "sonnet",
    "timezone": "America/Sao_Paulo"
  }'

Heartbeats

Heartbeats are periodic check-ins where agents review HEARTBEAT.md and take action if needed.

Configure per agent in openclaw.json:

{
  "agents": {
    "defaults": {
      "heartbeat": {
        "every": "30m",
        "target": "last",
        "activeHours": { "start": "08:00", "end": "22:00" }
      }
    }
  }
}

Edit HEARTBEAT.md in the agent’s workspace to define what it checks:

HEARTBEAT.md
# Heartbeat Checklist


- [ ] Check for new tasks assigned to me
- [ ] Review any blocked tasks and report status
- [ ] Update memory/ with significant findings
- [ ] Check if any deployments are pending review

When to use cron vs heartbeat

Use CaseCronHeartbeat
Exact timing requiredYesNo
Batch multiple checksNoYes
Needs session contextNoYes
Different model per taskYesNo
One-shot remindersYesNo

Next Steps