Skip to main content

Agent Skills

CLEO provides a comprehensive skills system for extending AI agent capabilities through protocol injection.
Architecture (v0.70.0+): CLEO uses a 2-tier universal subagent architecture. All spawns use cleo-subagent with skill-based protocol injection. Skill names like ct-research-agent identify protocols, not separate agent types.

Skill Taxonomy

Skills are organized by tier (execution order) and category (functional domain):

Tier 0: Orchestration

Human-in-the-Loop entry points that coordinate workflows.
SkillDescriptionTags
ct-orchestratorMulti-agent workflow coordinationworkflow, multi-agent, coordination

Tier 1: Planning

Decompose complex work into executable units.
SkillDescriptionTags
ct-epic-architectEpic decomposition and task planningplanning, architecture, task-management

Tier 2: Execution

Perform the actual work.
SkillDescriptionTags
ct-task-executorGeneral implementation workflowsexecution, implementation
ct-research-agentInformation gathering and analysisresearch, investigation, discovery
ct-spec-writerTechnical specifications and RFCsspecification, documentation, rfc
ct-test-writer-batsBATS test creationtesting, bats, integration
ct-library-implementer-bashBash library developmentimplementation, bash, library
ct-validatorValidation and compliance checksvalidation, compliance, audit
ct-dev-workflowRelease and deployment workflowsworkflow, git, release

Tier 3: Support

Specialized utilities and meta-skills.
SkillDescriptionTags
ct-documentorDocumentation orchestrationdocumentation, orchestration
ct-docs-lookupLibrary documentation via Context7documentation, libraries
ct-docs-writeDocumentation writing (CLEO style)documentation, writing
ct-docs-reviewDocumentation reviewdocumentation, review
ct-skill-creatorSkill development guideskills, creation, meta
ct-skill-lookupSkill discovery (SkillsMP)skills, discovery

Dispatch Matrix

Skills are automatically selected based on task context:

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"
}

By Keyword

KeywordsSkill
research, investigate, explorect-research-agent
epic, plan, decompose, architectct-epic-architect
implement, build, execute, createct-task-executor
test, bats, coveragect-test-writer-bats
doc, document, readme, guidect-documentor
spec, rfc, protocol, contractct-spec-writer
validate, verify, audit, compliancect-validator
release, version, publish, deployct-dev-workflow

Dispatch Flow

User Request: "research authentication methods"


skill_auto_dispatch("T1234")

        ├─1→ Check task labels
        ├─2→ Match keywords ("research" matches)
        ├─3→ skill_dispatch_by_keywords("research authentication")
        │         → returns "ct-research-agent"


skill_prepare_spawn("ct-research-agent", "T1234")


Spawn cleo-subagent with research protocol injection

Using Skills

Via CLI Command

cleo skills list                          # List installed skills
cleo skills search "query"                # Search local skills
cleo skills search "query" --mp           # Search marketplace (100K+)
cleo skills info ct-orchestrator          # Show skill details
cleo skills install @author/skill --mp    # Install from marketplace
cleo skills installed                     # List marketplace-installed skills
cleo skills update                        # Check for updates

Via Slash Command

/ct-orchestrator     # Invoke orchestrator
/ct-epic-architect   # Plan an epic
/ct-research-agent   # Research a topic

Programmatic Dispatch

# Auto-select skill based on task
protocol=$(skill_auto_dispatch "T1234")

# Prepare spawn with fully-resolved prompt
context=$(skill_prepare_spawn "$protocol" "T1234")

# Spawn cleo-subagent
Task tool: subagent_type="cleo-subagent", prompt=context.prompt

Documentation

Individual Skills

Tier 0-1: Coordination & Planning

Tier 2: Execution

Tier 3: Support