Documentation
Soul ID Docs
Everything you need to install agents, run multi-agent workflows, and contribute back to the catalog.
What is Soul ID?
Soul ID is a curated marketplace of production-ready OpenClaw agent bundles and multi-agent workflow patterns — available in English and Spanish.
Unlike repos that only ship a single prompt file, every agent here includes a complete workspace bundle — SOUL, IDENTITY, USER context, memory conventions, periodic tasks, tool config, and a bootstrap guide — ready to drop into OpenClaw and start working immediately.
Install an Agent
The fastest way to install an agent is directly from the Soul ID UI. Open any agent page, click Deploy in OpenClaw, and the install command is copied to your clipboard. Then paste it into OpenClaw.
One-line CLI install
You can also install any agent from your terminal using the install script:
# Install an agent (prompts you to select a workspace)
curl -fsSL https://raw.githubusercontent.com/cerealskill/openclaw-agents/main/install.sh \
| bash -s agent <slug> EN
# Example: install pedro-sre in English
curl -fsSL https://raw.githubusercontent.com/cerealskill/openclaw-agents/main/install.sh \
| bash -s agent pedro-sre EN
# Example: install in Spanish
curl -fsSL https://raw.githubusercontent.com/cerealskill/openclaw-agents/main/install.sh \
| bash -s agent pedro-sre ESInstall script parameters
| Parameter | Values | Description |
|---|---|---|
| 1st | agent | workflow | What to install |
| 2nd | <slug> | Agent or workflow slug, e.g. pedro-sre or incident-mode |
| 3rd | EN | ES | Language variant to install |
Manual install
Download the .tar.gz from any agent page, extract it, and copy the folder into your OpenClaw workspace directory (usually ~/.openclaw/workspace-<name>/agents/<slug>/).
After installing, open OpenClaw, select the workspace where you installed the agent, and start a new session. The agent loads its SOUL and AGENTS files automatically on startup.
Bundle Structure
Every agent bundle is a self-contained directory. When you install an agent, the following files are copied into your workspace. OpenClaw reads them in a defined order to build the agent's context window on session start.
agents/<lang>/<category>/<subcategory>/<slug>/
├── meta.yaml ← registry metadata
├── SOUL.md ← core identity (loaded first)
├── IDENTITY.md ← display name & avatar
├── USER.md ← your personal context
├── AGENTS.md ← workspace rules & memory conventions
├── HEARTBEAT.md ← periodic background tasks
├── TOOLS.md ← local tool config
└── BOOTSTRAP.md ← first-run setup guide| File | Purpose | Required |
|---|---|---|
| SOUL.md | Core identity — personality, values, tone, and behavioral rules. The most important file. | Yes |
| IDENTITY.md | Display name, creature type, emoji, and visual avatar metadata. | Yes |
| USER.md | Context about the person this agent works with (name, role, preferences). | Yes |
| AGENTS.md | Workspace operating rules, memory conventions, and session startup checklist. | Yes |
| HEARTBEAT.md | Periodic background tasks executed on a schedule (daily reviews, cleanups, etc.). | Optional |
| TOOLS.md | Local tool configuration, environment notes, and integration-specific settings. | Optional |
| BOOTSTRAP.md | First-run setup guide — onboarding instructions the agent follows on day one. | Optional |
| meta.yaml | Registry metadata: name, slug, category, tags, model preference, and version. | Yes |
Load order
OpenClaw loads bundle files in this priority order on session start: SOUL.md → IDENTITY.md → USER.md → AGENTS.md. The remaining files are loaded on-demand or on schedule.
Customizing your install
Edit USER.md with your name, role, timezone, and preferences. This is the most impactful personalization — the agent uses this context in every session. Edit TOOLS.md to add local environment details (API endpoints, repo paths, team contacts).
Using Workflows
Workflows define multi-agent orchestration patterns — how multiple agents coordinate, vote, escalate, and hand off work. Each workflow ships as an ORCHESTRATION.md (EN) or ORQUESTACION.md (ES) file you install into a workspace.
Install a workflow
# Install a workflow — the script prompts you to pick a workspace
curl -fsSL https://raw.githubusercontent.com/cerealskill/openclaw-agents/main/install.sh \
| bash -s workflow <slug> EN
# Example: install the incident-mode workflow
curl -fsSL https://raw.githubusercontent.com/cerealskill/openclaw-agents/main/install.sh \
| bash -s workflow incident-mode ENThe script copies the ORCHESTRATION.md file into your selected agent workspace. You can then reference it when instructing OpenClaw to coordinate multiple agents.
How to activate a workflow in OpenClaw
Workflows are instructions for the orchestrator agent — typically your main agent or a dedicated coordinator. To activate one:
Workflows are language-agnostic protocols — they describe coordination logic, not content. You can use an EN workflow with ES agents, or mix languages freely. The pattern is what matters.
Teams
A Team is a pre-configured roster of agents that work together toward a shared goal, optionally wired to a workflow pattern. Teams are a first-class concept in Soul ID — you can browse them in the catalog, deploy all members at once, and drop them straight into OpenClaw.
Each team ships as two files: a team.yaml that declares the roster and metadata, and a ROSTER.md that explains the objective, interaction model, and use cases in human-readable form.
team.yaml structure
teams/<lang>/<category>/<slug>/
├── team.yaml ← roster metadata
└── ROSTER.md ← human-readable interaction guide| Field | Description | Required |
|---|---|---|
| name | Display name of the team | Yes |
| slug | URL-safe identifier used in installs and registry | Yes |
| description | One-line summary of what the team does together | Yes |
| category | Team category (devops, development, marketing, security, …) | Yes |
| members | List of agent slugs with their role in this team | Yes |
| workflow | Workflow slug this team follows (e.g. incident-mode) | Optional |
| tags | Array of searchable tags | Optional |
| language | en or es | Yes |
| version | Semver string (1.0.0) | Yes |
ROSTER.md structure
The ROSTER.md is a markdown document that describes the team in detail. It follows a standard template with sections for Team, Objective, Members (table with agent, role, and responsibility), Interaction Model (ASCII flow diagram), When to Use, Use Cases, and Rules.
Deploying a team
Open any team page in the Soul ID catalog and click Deploy Team. This installs all member agents into your OpenClaw instance at once. You can also install members individually using the standard agent install command:
# Install each team member individually
curl -fsSL https://raw.githubusercontent.com/cerealskill/openclaw-agents/main/install.sh \
| bash -s agent <member-slug> ENActivating the team workflow
Most teams reference a workflow in their team.yaml. After installing all members, install the linked workflow into your coordinator agent workspace. Then instruct the coordinator: "Use the [workflow name] defined in ORCHESTRATION.md — the full team roster is in ROSTER.md."
Teams are declarative — they define who collaborates and how, but each member agent retains its full individual identity, SOUL, and expertise. A team is a coordination contract on top of existing agents.
Workflow Patterns Reference
34 orchestration patterns organized by use case. Each links to its full ORCHESTRATION.md.
🚨 Incident & Operations
🔄 Execution & Resilience
🧠 Decision & Review
⚡ Parallel & Distribution
🔬 Research & Knowledge
🤝 Collaboration
Contributing
All contributions are welcome. The easiest path is through the web UI — no git required.
Via the web form
Via GitHub PR
Fork the repo, create the bundle directory following the structure above, and open a Pull Request against main. The meta.yaml file is required for registry inclusion.
# Fork & clone
git clone https://github.com/cerealskill/openclaw-agents.git
cd openclaw-agents
# Create your agent bundle
mkdir -p agents/EN/<category>/<subcategory>/<your-slug>
# Add SOUL.md, IDENTITY.md, USER.md, AGENTS.md, meta.yaml
# Commit and push, then open a PR on GitHubThe meta.yaml must include at minimum: name, slug, description, category, subcategory, and version. The registry build script reads this to generate the catalog.
FAQ
What is OpenClaw?
OpenClaw is an AI agent orchestration platform. Agents on this site are designed specifically for OpenClaw workspaces — each bundle follows OpenClaw file conventions and loads automatically on session start.
Can I use these agents with other platforms?
The SOUL.md and IDENTITY.md files are plain markdown — you can read them into any LLM context. The other files (AGENTS.md, HEARTBEAT.md) are OpenClaw-specific conventions. SOUL.md alone is a useful system prompt for any platform.
What's the difference between an agent and a workflow?
An agent is a single AI persona with a defined identity, expertise, and operating rules. A workflow is an orchestration protocol that coordinates two or more agents — defining roles, protocols, escalation paths, and deliverables.
How do I customize an agent after installing it?
Edit USER.md with your personal context (name, role, preferences, working style). Edit TOOLS.md for your local environment. You can also edit SOUL.md directly to adjust the agent's personality, but keep the core identity section intact.
Are agents available in both English and Spanish?
Yes. Every agent and workflow ships in both EN and ES. The language toggle on the catalog switches the entire UI. When installing via CLI, pass EN or ES as the third argument.
How often is the catalog updated?
The registry is rebuilt on every merge to main. New agents appear in the catalog immediately after their PR is merged. The web app reads from the generated workflows-data.ts and agents-data.ts at build time.
