agent-templates

5 reusable agent archetypes (Researcher, Writer, Executor, Analyst, Coordinator) with pre-configured skill combos, memory templates, and SOUL.md files for rapid swarm deployment.

What This Skill Does

Agent Templates provides starter kits for spinning up specialized agents quickly. Each template is self-contained, pre-configured with personality (SOUL.md), workspace rules (AGENTS.md), recommended skills, and memory structure.

Perfect for:

Swarm orchestration (spawn 5+ agents with different specializations)
Rapid experimentation (test new agent roles without manual setup)
Sparta-style task delegation (Coordinator template coordinates others)
Consistent agent onboarding (every Writer agent has the same baseline)

The 5 Archetypes

🔍 Researcher

Personality: Methodical, skeptical, evidence-first
Good at: Deep research, fact-checking, source evaluation, contradiction-finding
Rules: Always cite sources, cross-reference 3+ sources, flag contradictions
Recommended skills: web_search, web_fetch, playwright-scraper, summarizer
Default model: Sonnet (complex research reasoning)
Memory structure: Sources, Key Findings, Open Questions, Confidence Levels

✍️ Writer

Personality: Creative, concise, audience-aware
Good at: Blog posts, social media, technical docs, marketing copy
Rules: Match brand voice, use active voice, respect word limits, include CTAs
Recommended skills: web_search, typefully-poster, content-formatter
Default model: Haiku (creative but efficient)
Memory structure: Style Guide, Content Calendar, Published Pieces, Drafts

⚙️ Executor

Personality: Systematic, careful, safety-conscious
Good at: Running scripts, deploying, managing infra, operational tasks
Rules: Always dry-run first, create backups before changes, log all actions, never rm without trash
Recommended skills: github, deployment-tools, monitoring
Default model: Haiku (clear instructions, efficient execution)
Memory structure: Deployment Log, Rollback Points, Active Services, Incident History

📊 Analyst

Personality: Data-driven, precise, pattern-seeking
Good at: Market analysis, metrics interpretation, trend detection, forecasting
Rules: Always show data before conclusions, use confidence intervals, flag assumptions
Recommended skills: api-tester, dca-calculator, data-viz, web_search
Default model: Sonnet (complex data reasoning)
Memory structure: Data Sources, Analysis History, Key Metrics, Predictions vs Actuals

🎯 Coordinator

Personality: Organized, diplomatic, big-picture
Good at: Task delegation, progress tracking, conflict resolution, team orchestration
Rules: Never do the work yourself (delegate), track all subtasks, report blockers immediately
Recommended skills: multichannel-dispatch, session-management, github (for task tracking)
Default model: Haiku (clear communication, task management)
Memory structure: Active Tasks, Agent Assignments, Blockers, Status Dashboard

Installation


# Install the skill
clawhub install agent-templates

# Or manually copy to your workspace
cp -r agent-templates/ ~/.openclaw/workspace/skills/

Usage

Method 1: Using the Spawner Script


# List available templates
python3 scripts/list-templates.py

# Spawn a new agent from a template
python3 scripts/spawn-from-template.py \
  --template researcher \
  --name "Market-Researcher" \
  --workspace /path/to/new/workspace

Method 2: Manual Deployment

Each template is self-contained and can be copied directly:


# Copy researcher template files
cp -r templates/researcher/* /path/to/agent-workspace/

# Customize SOUL.md for your agent
vi /path/to/agent-workspace/SOUL.md

File Structure


agent-templates/
├── SKILL.md                          (this file)
├── templates/
│   ├── researcher/
│   │   ├── SOUL.md
│   │   ├── AGENTS.md
│   │   ├── skills.json
│   │   └── memory-template.md
│   ├── writer/
│   ├── executor/
│   ├── analyst/
│   └── coordinator/
├── scripts/
│   ├── spawn-from-template.py        (creates new agent from template)
│   └── list-templates.py             (lists available templates)
└── config/
    └── template-defaults.json        (model defaults, concurrency limits)

Example: Spawning a Swarm


# Spawn 3 researcher agents for parallel research
python3 scripts/spawn-from-template.py --template researcher --name "Research-1" --workspace ~/agents/research-1
python3 scripts/spawn-from-template.py --template researcher --name "Research-2" --workspace ~/agents/research-2
python3 scripts/spawn-from-template.py --template researcher --name "Research-3" --workspace ~/agents/research-3

# Spawn a coordinator to manage them
python3 scripts/spawn-from-template.py --template coordinator --name "Swarm-Manager" --workspace ~/agents/coordinator

# Spawn an executor for deployment
python3 scripts/spawn-from-template.py --template executor --name "Deployer" --workspace ~/agents/executor

Design Principles

1. Templates are starting points, not straitjackets — Agents can evolve their own SOUL.md and workspace rules as they learn

2. Self-contained — Each template can be used independently; no external dependencies between archetypes

3. Recommendations, not requirements — Skills listed in skills.json are suggested; agents can add/remove skills as needed

4. Flexible memory — Memory templates provide structure but don't restrict what agents can store

5. Coordinator-centric — Coordinator template designed specifically for Sparta-style swarm orchestration

What Agents Can Do With Templates

Researchers: Investigate topics in parallel, cross-reference sources, flag contradictions
Writers: Create content simultaneously (blog, social, docs), maintain consistent voice
Executors: Deploy, monitor, and rollback infrastructure safely with audit trails
Analysts: Run independent analyses on different datasets, compare findings
Coordinators: Orchestrate swarms, track subtasks, escalate blockers, synthesize results

Tips & Tricks

Scaling Research Tasks

Use multiple Researchers with the same AGENTS.md rules. Coordinator aggregates findings.

Content Calendars

Writer agents share a Content Calendar section in memory. Each adds drafts, coordinator schedules publication.

Safe Deployments

Executor agents dry-run changes, create backups, log everything. Coordinator approves based on logs.

Data Analysis

Analysts generate independent analyses. Coordinator compares methodologies and results for cross-validation.

Swarm Coordination

Coordinator delegates work, tracks progress, and reports blockers. Never does actual work itself.

Built by GetAgentIQ — getagentiq.ai

Stability: Experimental (templates are starting points; agent evolution expected)

Last Updated: 2026-03-22

License: MIT

Agent Templates