Presets

Presets

System prompts for ling/drone specialization. Presets define agent behavior, constraints, and methodology.

Overview

Presets are markdown files in the presets/ directory that get migrated to Chroma for semantic search. When spawning agents, you can combine multiple presets to create specialized workers.

┌─────────────────────────────────────────────────────────┐
│  presets/*.md  →  preset_migrate()  →  Chroma Vector DB │
│                                                         │
│  swarm_spawn(presets: ["tdd", "clarity"])              │
│       ↓                                                 │
│  Agent receives combined system prompts                 │
└─────────────────────────────────────────────────────────┘

Available Presets (38)

The tables below cover the main categories. Run preset list for the authoritative current set, or preset search "<describe the job>" to find one by intent. See also Agents-and-Skills for how presets differ from agent definitions.

Core Agent Types

Preset	Purpose
ling	Coordinator agent with hivemind integration
ling-worker	Token-optimized leaf agent
hivemind	Event-driven swarm coordination
hivemind-master	Can spawn lings and manage tasks
drone-worker	Diff-based safe agent (propose_diff only)
minimal	Bare minimum preset

Methodology Presets

Preset	Purpose
tdd	Test-Driven Development (Red-Green-Refactor)
ddd	Domain-Driven Design patterns
solid	SOLID principles enforcement
clarity	CLARITY framework for Go systems
tao	Truthful Adaptive Ordering

Role Presets

Preset	Purpose
architect	System design and architecture
analyzer	Code analysis and investigation
reviewer	Code review specialist
documenter	Documentation writer
fixer	Bug fixing specialist
refactorer	Refactoring specialist
researcher	Codebase exploration
tester	Test runner
verifier	Verification specialist
compliance	Compliance checking
metrics	Metrics and observability

Coordination Presets

Preset	Purpose
task-coordinator	Task distribution
wave-coordinator	Self-healing orchestration
hive-master	Swarm prompt handler

Workflow Presets

Preset	Purpose
mcp-first	Prefer MCP tools over native tools
planner-nih	NIH audit planning

MCP Tools

Tool	Purpose	Example
preset_list	List all presets with metadata	preset_list()
preset_get(name)	Get full preset content	preset_get(name: "tdd")
preset_search(query)	Semantic search by description	preset_search(query: "testing methodology")
preset_migrate(directory)	Migrate .md files to Chroma	preset_migrate(directory: "/path/to/project")
preset_add(name, content)	Add custom preset	preset_add(name: "my-preset", content: "...")
preset_delete(name)	Remove preset from Chroma	preset_delete(name: "my-preset")
preset_status	Get Chroma config and counts	preset_status()

Preset Lifecycle

1. Edit presets/*.md files
2. Run preset_migrate(directory: "/path/to/hive-mcp")
3. Chroma updated with new embeddings
4. Semantic search now finds updated content

Migration Example

;; After editing presets/my-new-preset.md
preset_migrate(directory: "/home/user/hive-mcp")
;; → {:migrated 27, :errors 0}

;; Verify
preset_list()
;; → Shows all presets including new one

Using Presets

With swarm_spawn

;; Single preset
swarm_spawn(name: "test-runner", presets: ["tdd"])

;; Multiple presets (combined)
swarm_spawn(name: "quality-checker", presets: ["tdd", "reviewer", "solid"])

;; With role shorthand
swarm_spawn(name: "docs-writer", role: "documenter")

Preset Combination

When multiple presets are specified, their instructions are concatenated. Order matters - later presets can override earlier ones.

presets: ["ling", "tdd", "clarity"]
         ↓
System prompt = ling.md + tdd.md + clarity.md

With delegate_drone

;; Drones use presets too
delegate_drone(
  task: "Add docstrings to api.clj",
  preset: "drone-worker"  ;; Default
)

Creating Custom Presets

File Format

Presets are markdown files with optional YAML-style metadata in the title:

# My Custom Preset

You are a specialized agent for [purpose].

## Rules

- Rule 1
- Rule 2

## Tools to Use

| Tool | When |
|------|------|
| `tool_name` | Description |

## Anti-Patterns

- Don't do X
- Avoid Y

Category Options

When adding via preset_add, use one of these categories:

Category	Use For
testing	Test-related presets
coding	Implementation presets
architecture	Design/structure presets
coordination	Multi-agent presets
workflow	Process presets
custom	Everything else

Adding Custom Preset

preset_add(
  name: "my-project-conventions",
  content: "# My Project Conventions\n\n## Naming\n- Use snake_case for files\n- Use kebab-case for namespaces\n\n## Testing\n- Every public fn needs a test",
  category: "custom",
  tags: ["project-specific", "conventions"]
)

Preset Examples

TDD Preset (Methodology)

# TDD (Test-Driven Development) Preset

You follow strict **Red-Green-Refactor** discipline.

## The TDD Cycle

1. **RED**: Write a failing test that expresses minimal new behavior
2. **GREEN**: Make it pass with the simplest possible change
3. **REFACTOR**: Remove duplication, clarify names, keep tests green

## Rules

- Keep cycles **small** (<5 minutes ideally)
- Write the **test first**, always
- Only write **production code to make a failing test pass**

Drone Worker Preset (Agent Type)

The drone-worker preset is critical for safe file operations:

• Cannot use file_write, file_edit, or direct mutations
• Must use propose_diff for all changes
• Must run kondo_lint before proposing Clojure changes
• Hivemind reviews and applies/rejects proposed diffs

Semantic Search

Find presets by describing what you need:

;; Find testing-related presets
preset_search(query: "test driven development methodology")
;; → [tdd, tester, reviewer]

;; Find coordination presets
preset_search(query: "multi-agent orchestration")
;; → [hivemind, wave-coordinator, task-coordinator]

;; Find architecture presets
preset_search(query: "domain driven design patterns")
;; → [ddd, architect, solid]

Best Practices

1. Start with ling - Base preset for coordinator agents
2. Add methodology - tdd, solid, clarity as needed
3. Add role - reviewer, documenter, etc. for specialization
4. Keep presets focused - One concern per preset
5. Use semantic search - Find existing presets before creating new ones
6. Migrate after edits - Run preset_migrate after changing .md files

Troubleshooting

Preset not found after edit

;; Re-migrate to sync Chroma
preset_migrate(directory: "/path/to/hive-mcp")

Semantic search returns wrong results

;; Check preset status
preset_status()
;; Verify Chroma is configured and count matches file count

Custom preset not persisting

Custom presets added via preset_add are stored in Chroma only. To make them permanent, create a .md file in presets/ and run preset_migrate.

Related Pages

• Session-Continuity - Using presets with /wrap and /catchup
• Infrastructure-Setup - Chroma configuration for semantic search
• Seed-Memories - Bootstrap knowledge that complements presets