AI Skill Templates: Structured, Governed, Ready to Run

118+ SKILL.md templates across 17 categories. Each template is a single atomic AI task with model recommendations, token estimates, and Croms scores.

Looking for multi-skill workflows? Browse Workflow Templates instead.

What Is an AI Skill Template?

A skill is one atomic AI task. It does one thing, with one model, and produces one output. Every skill template is structured as a SKILL.md file — a portable, human-readable specification that any platform can parse and execute.

Model recommendation
Input contract
Output contract
Ordered steps
Token estimates
Croms™ baseline
Caching notes
Workflow position

17 Skill Categories

Every skill template belongs to a category that defines its function within a workflow graph.

Research

Information gathering, source analysis, and knowledge synthesis tasks.

SERP Sampling
Entity Extraction
Key Finding Extraction

Content

Content creation, structuring, and formatting tasks.

Outline Generator
Brief Formatter
Final Report Generator

Development

Code analysis, documentation, and software engineering tasks.

Chunking Plan
Docstring Generation
Embedding Plan

Technical

Technical analysis, auditing, and diagnostic tasks.

Issue Classification
Priority Scoring
Fix Recommendation Draft

Operations

Process automation, triage, and operational decision tasks.

Intent Classification
Routing Recommendation
Response Draft

Marketing

Campaign creation, competitive analysis, and messaging tasks.

Social Post Variants
Positioning Extraction
Feature Comparison

Accessibility

WCAG compliance, keyboard testing, and accessibility audit tasks.

Automated Scan
WCAG Severity Scoring
Issue Prioritization

CMS

Content management, publishing QA, and digital asset governance tasks.

Content QA Checks
DAM Asset Validation
Governance Gate

Data

Data loading, ingestion, and preparation tasks.

Document Ingestion
Data Loader
Page Crawler

Processing

Data transformation, aggregation, and pipeline processing tasks.

Chunk Processor
Result Aggregator
Output Merger

Planning

Task planning, decomposition, and strategy formulation tasks.

Query Planner
Task Decomposer
Step Decomposition

Validation

Quality assurance, fact-checking, and output verification tasks.

Fact Checker
Reviewer Agent
Validation Checks

Analysis

Deep analysis, pattern detection, and insight extraction tasks.

Local Analyzer
Content Gap Finder
Competitive Analyzer

Storage

Data persistence, indexing, and retrieval infrastructure tasks.

Qdrant Index
Embedding Generation
Vector Store

Startup

Founder content, launch planning, and growth automation tasks.

LinkedIn Post from Raw Idea
Investor One-Pager Writer
Cold Investor Email

Governance

Compliance, review enforcement, and governance checkpoint tasks.

Compliance Check
Review Cycle Gate
Approval Requirements

Workflow Position

Skills tagged by their position in a multi-skill workflow graph.

Entry Point
Mid-Chain Router
Terminal Output

What's Inside a SKILL.md File

Every SKILL.md template uses YAML frontmatter fields that define the skill's behavior, cost profile, and platform compatibility.

FieldDescription
nameUnique skill identifier. Used as the directory name on export.
versionSemantic version (e.g. 1.0.0). Required for registry submission on OpenClaw.
modelRecommended model (e.g. claude-sonnet-4-20250514). Cromus maps this to the model registry for cost simulation.
complexityTask complexity tier: eco, cost, balanced, or quality. Drives model selection in the cost simulator.
tokensEstimated input/output token counts. Used for per-run and monthly cost projections.
cromsBaseline Croms™ score. Measures preventable waste before the skill has ever been executed.
inputsRequired input data, formats, and prerequisites. Defines what the skill expects to receive.
stepsOrdered list of prompt instructions. Each step is one LLM call or tool invocation.
outputsExpected output format, schema, or template. Defines the skill's delivery contract.
workflow_positionWhere this skill sits in a multi-skill graph: entry, mid-chain, or terminal.
compatibilityPlatform compatibility notes. Used by AgentSkills.io for registry filtering.

Platform Compatibility

SKILL.md templates are portable across four platforms. Each platform supports a different subset of frontmatter fields — unsupported fields are stripped on export.

PlatformSupported FieldsStripped on Export
Cromus NativeAll fields, interoperability.json, ops-report.mdNone — all fields preserved
Replit Agentname, description, enabled, model, stepslicense, compatibility, metadata, allowed-tools
AgentSkills.ioname, description, model, steps, compatibility, metadata, license, allowed-toolsenabled (not part of spec)
OpenClawname, version, description, model, steps, security, license, metadataenabled, compatibility (AgentSkills.io field)

Frequently Asked Questions

What is an AI skill template?

How do I import a skill template into my project?

How is the Croms™ baseline calculated for a skill template?

What is the difference between a skill template and a workflow template?

Need connected multi-skill graphs? Browse Workflow Templates. Want to install an entire capability area at once? Grab a Skill Pack.

Browse all 118+ skill templates

Sign in to explore every template in the Skill Studio, simulate costs across model tiers, and start building.

Start Building

What is an AI skill template?

A skill template is a structured SKILL.md file that defines one atomic AI task — a single prompt, a single model, a single output. It includes the model recommendation, input/output contracts, step-by-step instructions, token estimates, and a baseline Croms™ score. Skills are the building blocks inside larger workflow templates.

How do I import a skill template into my project?

Sign in to Cromus, go to the Skill Studio, and browse the template library. Select a template, review the frontmatter fields and instructions, then click 'Use Template' to add it to your workspace. You can also export any SKILL.md to Replit Agent, AgentSkills.io, or OpenClaw format using the Platform Selector.

How is the Croms™ baseline calculated for a skill template?

The Croms baseline is calculated by analyzing four factors: cost waste (35% weight) from model-tier mismatches, latency overhead (25%) from serial execution patterns, failure risk (25%) from missing fallback or error handling, and structural gaps (15%) from absent context requirements or output contracts. A lower baseline score means less preventable waste.

What is the difference between a skill template and a workflow template?

A skill template is a single atomic task — one prompt, one model, one output. A workflow template is a connected graph of multiple skills that automates an end-to-end process. Skills are the nodes; workflows define the edges, branching, and data flow between them. Browse workflow templates at /ai-workflow-templates.