Manifest Schema

Skill registry format and dispatch matrix specification

Manifest Schema

The skills/manifest.json is the source of truth for skill metadata, dispatch configuration, and taxonomy.

Schema Overview

{
  "$schema": "https://cleo-dev.com/schemas/v1/skills-manifest.schema.json",
  "_meta": {
    "schemaVersion": "2.1.0",
    "lastUpdated": "2026-01-27",
    "totalSkills": 15,
    "architectureNote": "Universal Subagent Architecture..."
  },
  "dispatch_matrix": { ... },
  "skills": [ ... ]
}

Top-Level Structure

Field	Type	Description
$schema	string	JSON Schema reference
_meta	object	Manifest metadata
dispatch_matrix	object	Skill selection rules
skills	array	Skill definitions

_meta Object

{
  "_meta": {
    "schemaVersion": "2.1.0",
    "lastUpdated": "2026-01-27",
    "totalSkills": 15,
    "architectureNote": "Universal Subagent Architecture: All spawns use single agent type 'cleo-subagent' with skill/protocol injection."
  }
}

Field	Type	Description
schemaVersion	string	Manifest schema version (semver)
lastUpdated	string	ISO date of last update
totalSkills	number	Count of skills in manifest
architectureNote	string	Architecture documentation

Dispatch Matrix

The dispatch matrix defines how skills are automatically selected:

by_task_type

Maps CLEO task types to skills:

{
  "dispatch_matrix": {
    "by_task_type": {
      "research": "ct-research-agent",
      "planning": "ct-epic-architect",
      "implementation": "ct-task-executor",
      "testing": "ct-test-writer-bats",
      "documentation": "ct-documentor",
      "specification": "ct-spec-writer",
      "validation": "ct-validator",
      "bash-library": "ct-library-implementer-bash",
      "release": "ct-dev-workflow",
      "consensus": "ct-validator",
      "contribution": "ct-task-executor"
    }
  }
}

by_keyword

Maps regex patterns to skills:

{
  "dispatch_matrix": {
    "by_keyword": {
      "research|investigate|explore|discover": "ct-research-agent",
      "epic|plan|decompose|architect": "ct-epic-architect",
      "implement|build|execute|create": "ct-task-executor",
      "test|bats|coverage|integration": "ct-test-writer-bats",
      "doc|document|readme|guide": "ct-documentor",
      "spec|rfc|protocol|contract": "ct-spec-writer",
      "validate|verify|audit|compliance|consensus|vote": "ct-validator",
      "lib/|bash|shell|function": "ct-library-implementer-bash",
      "release|version|publish|deploy|ship|changelog": "ct-dev-workflow"
    }
  }
}

by_protocol

Maps protocol files to skills:

{
  "dispatch_matrix": {
    "by_protocol": {
      "research": "ct-research-agent",
      "consensus": "ct-validator",
      "specification": "ct-spec-writer",
      "decomposition": "ct-epic-architect",
      "implementation": "ct-task-executor",
      "contribution": "ct-task-executor",
      "release": "ct-dev-workflow"
    }
  }
}

Skill Definition

Each skill in the skills array follows this structure:

{
  "name": "ct-research-agent",
  "version": "1.0.0",
  "description": "Information gathering and analysis skill...",
  "path": "skills/ct-research-agent",
  "tags": ["research", "investigation", "discovery"],
  "status": "active",
  "tier": 2,
  "token_budget": 5000,
  "references": [
    "skills/ct-research-agent/SKILL.md",
    "skills/_shared/subagent-protocol-base.md"
  ],
  "capabilities": {
    "inputs": ["TASK_ID", "EPIC_ID", "SESSION_ID"],
    "outputs": ["research-output", "manifest-entry"],
    "dependencies": [],
    "dispatch_triggers": ["research", "investigate"]
  }
}

Field Reference

Field	Type	Required	Description
name	string	✅	Unique skill identifier (ct-* prefix)
version	string	✅	Skill version (semver)
description	string	✅	Human-readable description
path	string	✅	Directory path relative to skills/
tags	array	✅	Category tags for filtering
status	enum	✅	active, deprecated, experimental
tier	number	✅	0-3 hierarchy level
token_budget	number	❌	Recommended context budget
references	array	❌	Supporting documentation paths
capabilities	object	❌	Detailed capability specification

Capabilities Object

{
  "capabilities": {
    "inputs": ["TASK_ID", "EPIC_ID", "SESSION_ID"],
    "outputs": ["research-output", "manifest-entry"],
    "dependencies": ["ct-research-agent"],
    "dispatch_triggers": [
      "orchestrate",
      "coordinate agents",
      "multi-agent workflow"
    ]
  }
}

Field	Type	Description
inputs	array	Required token variables
outputs	array	Types of output produced
dependencies	array	Other skills this skill may invoke
dispatch_triggers	array	Keywords/phrases that trigger this skill

Tier Hierarchy

Tier	Name	Purpose	Skills
0	Orchestration	HITL entry point	ct-orchestrator
1	Planning	Decomposition	ct-epic-architect
2	Execution	Implementation	ct-task-executor, ct-research-agent, ct-spec-writer, ct-test-writer-bats, ct-library-implementer-bash, ct-validator, ct-dev-workflow
3	Support	Utilities	ct-documentor, ct-docs-lookup, ct-docs-write, ct-docs-review, ct-skill-creator, ct-skill-lookup

Tags Taxonomy

Standard tags for skill categorization:

Category	Tags
Workflow	workflow, multi-agent, coordination
Planning	planning, architecture, task-management
Execution	execution, implementation
Research	research, investigation, discovery
Testing	testing, bats, integration
Documentation	documentation, writing, review
Specification	specification, rfc
Validation	validation, compliance, audit
Skills	skills, creation, discovery

Querying the Manifest

List Skills by Tier

cat skills/manifest.json | jq '.skills | map(select(.tier == 2))'

List Skills by Tag

cat skills/manifest.json | jq '.skills | map(select(.tags | contains(["documentation"])))'

Get Dispatch Mapping

cat skills/manifest.json | jq '.dispatch_matrix.by_task_type'

Check Skill Status

cat skills/manifest.json | jq '.skills | map(select(.status == "active")) | length'

Validation

Validate manifest against schema:

cleo skills validate manifest

SkillsMP Integration

External skills from SkillsMP are stored in skills/mp/ with their own manifest:

skills/
├── ct-orchestrator/        ← Built-in (Tier 0-3)
├── ...
├── mp/                     ← SkillsMP external
│   ├── code-review/
│   ├── security-audit/
│   └── manifest.json       ← External skill metadata
└── manifest.json           ← Primary (merges with mp/)

The primary manifest can reference external skills:

{
  "skills": [
    {
      "name": "mp-code-review",
      "path": "skills/mp/code-review",
      "external": true,
      "source": "skillsmp",
      "scopedName": "@wshobson/code-review"
    }
  ]
}

Schema Reference

Full JSON Schema: schemas/skills-manifest.schema.json

cat schemas/skills-manifest.schema.json | jq '.properties'