Preset Agents

Wunderland ships with 8 agent presets and 3 deployment templates that provide ready-to-use configurations for common use cases. Presets define the agent's personality, security posture, suggested skills, and recommended channels -- so you can go from zero to a working agent in one command.

Using Presets via CLI

The fastest way to create a new agent from a preset:

wunderland init my-agent --preset research-assistant

This scaffolds a project directory with agent.config.json, PERSONA.md, and .env.example pre-populated from the preset.

You can combine a preset with a security tier override:

wunderland init my-agent --preset creative-writer --security-tier permissive

To list all available presets:

wunderland list-presets

Available Agent Presets

research-assistant

Thorough researcher with analytical focus. High conscientiousness and openness drive systematic, deep-dive research behavior.

Trait	Value	Security Tier	Suggested Skills
Honesty	0.90	balanced	web-search, summarize, github
Emotionality	0.30
Extraversion	0.40	Channels
Agreeableness	0.70	webchat, slack
Conscientiousness	0.95
Openness	0.85

---

customer-support

Patient, empathetic support specialist. Very high agreeableness and elevated emotionality produce warm, accommodating interactions. Uses strict security to protect customer data.

Trait	Value	Security Tier	Suggested Skills
Honesty	0.80	strict	healthcheck
Emotionality	0.70
Extraversion	0.60	Channels
Agreeableness	0.95	webchat, telegram, whatsapp, discord
Conscientiousness	0.85
Openness	0.50

---

creative-writer

Imaginative storyteller and content creator. Near-maximum openness (0.98) combined with high emotionality and moderate conscientiousness encourages experimental, expressive output.

Trait	Value	Security Tier	Suggested Skills
Honesty	0.70	balanced	summarize, image-gen
Emotionality	0.80
Extraversion	0.70	Channels
Agreeableness	0.60	webchat
Conscientiousness	0.50
Openness	0.98

---

code-reviewer

Precise, detail-oriented code analyst. Near-maximum conscientiousness (0.98) and high honesty (0.95) produce thorough, no-nonsense code reviews. Low emotionality keeps feedback objective.

Trait	Value	Security Tier	Suggested Skills
Honesty	0.95	strict	coding-agent, github
Emotionality	0.20
Extraversion	0.30	Channels
Agreeableness	0.50	webchat, slack, discord
Conscientiousness	0.98
Openness	0.70

---

data-analyst

Systematic data interpreter and visualizer. High conscientiousness and openness balanced with low emotionality for objective, thorough analysis.

Trait	Value	Security Tier	Suggested Skills
Honesty	0.90	balanced	summarize, coding-agent
Emotionality	0.20
Extraversion	0.40	Channels
Agreeableness	0.60	webchat, slack
Conscientiousness	0.90
Openness	0.80

---

security-auditor

Vigilant security-focused analyst. Near-maximum honesty (0.98) and conscientiousness (0.99) paired with the paranoid security tier. Low agreeableness means it will not hesitate to flag issues.

Trait	Value	Security Tier	Suggested Skills
Honesty	0.98	paranoid	coding-agent, github, healthcheck
Emotionality	0.15
Extraversion	0.25	Channels
Agreeableness	0.30	webchat
Conscientiousness	0.99
Openness	0.60

---

devops-assistant

Infrastructure and deployment specialist. Balanced extraversion with high conscientiousness for clear, reliable operational guidance. Uses strict security for production safety.

Trait	Value	Security Tier	Suggested Skills
Honesty	0.85	strict	healthcheck, coding-agent, github
Emotionality	0.20
Extraversion	0.50	Channels
Agreeableness	0.60	slack, discord, webchat
Conscientiousness	0.90
Openness	0.75

---

personal-assistant

Friendly, organized daily helper. High extraversion and agreeableness create a warm, proactive assistant. Suggested channels include the popular mobile messaging platforms.

Trait	Value	Security Tier	Suggested Skills
Honesty	0.80	balanced	weather, apple-notes, apple-reminders, summarize
Emotionality	0.60
Extraversion	0.75	Channels
Agreeableness	0.85	telegram, whatsapp, webchat
Conscientiousness	0.80
Openness	0.70

---

Deployment Templates

Templates provide full project configurations including security profiles, inference defaults, and step-up authorization rules. They are more comprehensive than agent presets.

minimal

Bare-bones configuration for local development and experimentation.

• Security tier: permissive
• All security layers disabled
• Inference: Ollama with llama3 (local)
• Channels: webchat only
• Step-up auth: Tier 1 (all autonomous), 10-minute timeout

wunderland init my-dev-agent --template minimal

standard

Balanced configuration suitable for most use cases. Enables input classification and dual-LLM auditing with moderate security.

• Security tier: balanced
• PreLLM classifier and DualLLM auditor enabled
• Inference: Ollama with dolphin-llama3:8b
• Channels: webchat, telegram, discord
• Skills: web-search, summarize, healthcheck
• Step-up auth: Tier 1 default, Tier 2 for data/API/communication, Tier 3 for financial/system

wunderland init my-agent --template standard

enterprise

Full security configuration for production and compliance-sensitive environments.

• Security tier: strict
• All three security layers enabled (including output signing)
• Inference: OpenAI gpt-4o with gpt-4o-mini router
• Channels: webchat, slack, teams, discord, telegram, whatsapp, email
• Skills: web-search, summarize, healthcheck, coding-agent, github
• Step-up auth: Tier 2 default, Tier 3 for data modification/financial/system
• Escalation triggers for high-value transactions, sensitive data, and irreversible actions
• Audit logging enabled (90-day retention)

wunderland init my-enterprise-agent --template enterprise

Preset File Structure

Each preset lives in a folder under presets/agents/<preset-id>/:

presets/agents/research-assistant/
  agent.config.json   # HEXACO traits, security tier, skills, channels
  PERSONA.md          # System prompt / personality description

Templates are single JSON files under presets/templates/:

presets/templates/
  minimal.json
  standard.json
  enterprise.json

Loading Presets Programmatically

import { PresetLoader } from 'wunderland/advanced';

const loader = new PresetLoader();

// List all presets
const presets = loader.listPresets();
for (const p of presets) {
  console.log(`${p.id}: ${p.description} (${p.securityTier})`);
}

// Load a specific preset
const researcher = loader.loadPreset('research-assistant');
console.log(researcher.hexacoTraits);
console.log(researcher.persona); // Contents of PERSONA.md

// Load a template
const enterprise = loader.loadTemplate('enterprise');
console.log(enterprise.securityTier); // "strict"

// Get all known preset IDs (static, no filesystem)
const ids = PresetLoader.getPresetIds();
// ['research-assistant', 'customer-support', 'creative-writer', ...]

Creating Custom Presets

To create your own preset:

1. Create a directory under presets/agents/:

mkdir -p presets/agents/my-custom-agent

2. Create agent.config.json:

{
  "name": "My Custom Agent",
  "description": "A domain-specific agent for my use case",
  "hexacoTraits": {
    "honesty": 0.85,
    "emotionality": 0.4,
    "extraversion": 0.6,
    "agreeableness": 0.7,
    "conscientiousness": 0.8,
    "openness": 0.75
  },
  "securityTier": "balanced",
  "suggestedSkills": ["web-search", "summarize"],
  "suggestedChannels": ["webchat", "slack"]
}

3. Create PERSONA.md with the agent's system prompt and behavioral guidelines.

4. Use it:

wunderland init my-project --preset my-custom-agent

Recommended Skills by Preset

Each agent preset declares suggestedSkills that are automatically loaded from the curated skills registry when you run wunderland start or wunderland chat in a project scaffolded from that preset.

Preset	Security	Skills
research-assistant	balanced	web-search, summarize, github
customer-support	strict	healthcheck
creative-writer	balanced	summarize, image-gen
code-reviewer	strict	coding-agent, github
data-analyst	balanced	summarize, coding-agent
security-auditor	paranoid	coding-agent, github, healthcheck
devops-assistant	strict	healthcheck, coding-agent, github
personal-assistant	balanced	weather, apple-notes, apple-reminders, summarize

How auto-loading works

1. wunderland init my-agent --preset research-assistant writes "skills": ["web-search", "summarize", "github"] into agent.config.json.
2. On wunderland start or wunderland chat, the CLI reads the skills array and calls resolveSkillsByNames() from the @framers/agentos-skills-registry package.
3. The resolver validates each skill name against the curated catalog, builds a prompt snapshot, and merges it into the system prompt alongside any directory-based skills.

You can modify the skills array in agent.config.json at any time to add or remove skills:

{
  "skills": ["web-search", "summarize", "github", "obsidian"]
}

Exporting and Importing Agents

Agents can be exported as portable JSON manifests for sharing, backup, or migration:

# Export current agent
wunderland export

# Export to a specific file
wunderland export -o my-agent-backup.json

# Import an agent from a manifest
wunderland import agent.manifest.json --dir ./my-imported-agent

The manifest includes HEXACO traits, security configuration, skills, channels, persona, and an optional integrity hash for sealed agents. Importing a sealed agent creates an unsealed copy. See Agent Serialization for the full manifest format.

Related

• Creating Agents -- full seed configuration reference
• HEXACO Personality -- personality trait system
• Security Tiers -- tier configuration details
• Skills System -- skill registry and management
• Agent Serialization -- export/import manifest format