The Agent Skills Marketplace
for AI Behavioral Protocols
ClawMerchants is the first marketplace for SKILL.md agent behavioral protocols. AI agents buy skills — expert roles, workflows, and decision frameworks — and load them at runtime. No fine-tuning required.
What Is SKILL.md?
SKILL.md is an emerging protocol for defining AI agent behavioral specifications as structured Markdown documents. Instead of trying to encode expert behavior into model weights through expensive fine-tuning, SKILL.md lets you express a complete expert role as a text document that any capable LLM can load and embody.
A SKILL.md file contains everything needed for an AI agent to perform at expert level in a specific domain:
Identity & Mindset
How the agent thinks, what it values, and what expert perspective it brings to problems.
Workflow Process
Step-by-step methodology the agent follows — from discovery through delivery — ensuring consistent, expert-level output.
Decision Rules
Critical rules and constraints that prevent common mistakes and ensure quality. The things an expert knows not to do.
Success Criteria
Measurable outcomes that define what "done well" looks like, enabling agents to self-assess output quality.
This is what makes the agent skills protocol so powerful: skills are composable, transferable, and instantly deployable. An agent that needs to review code buys the Code Review Skill and immediately operates at senior engineer level — no training, no configuration.
Available Skills
ClawMerchants hosts skills authored by domain experts and accessed by AI agents via x402 micropayments. Each skill is a complete behavioral protocol delivered as text.
A senior software engineer's complete code review methodology. Covers correctness, security, performance, and maintainability. Opinionated about what matters (bugs, vulnerabilities) and what doesn't (style preferences, over-engineering).
- Security vulnerability detection (OWASP Top 10 awareness)
- Performance and complexity analysis with Big-O reasoning
- Constructive feedback framing — actionable, not just critical
- Prioritized findings: blockers vs. suggestions vs. nitpicks
- Language-agnostic core with pattern recognition across stacks
An application security engineer's vulnerability assessment framework. OWASP-aligned, covers web application security, API security, authentication flows, injection attacks, and secrets management. Produces structured findings with CVSS-style severity ratings.
- OWASP Top 10 and beyond — covers common and uncommon attack vectors
- API security: authentication, authorization, rate limiting, data exposure
- Injection detection: SQL, command, LDAP, template, and more
- Cryptography review: weak algorithms, key management, entropy issues
- Structured findings: Critical / High / Medium / Low / Informational
More skills are added regularly. Interested in listing your own skill? Register as a provider to contribute skills to the marketplace.
How Agents Use Skills
The skill acquisition flow is identical to data asset access — the x402 payment protocol handles both. The difference is what gets returned: skills deliver SKILL.md text instead of JSON data.
-
Discover Skills via MCP
An agent calls
list_assetson the MCP server, filtering by categoryagent-skills. It receives skill IDs, descriptions, and prices. Theget_assettool provides full skill details before committing to purchase. -
Request the Skill (Receive 402)
The agent calls
GET https://clawmerchants.com/v1/data/{skill-id}. It receives a 402 Payment Required response with USDC payment instructions — identical to data assets. -
Pay and Receive Skill
After paying with USDC on Base L2, the agent retries the request and receives the full SKILL.md text — the complete behavioral protocol for that skill.
-
Load and Execute
The agent incorporates the SKILL.md text into its context — as a system prompt, a user message, or a mid-task role assignment. It then operates as the specified expert for the duration of the task.
Why Skills Marketplaces Are the Future
The current approach to specializing AI agents — fine-tuning, RLHF, custom training runs — is expensive, slow, and fragile. A skills marketplace enables an entirely different model:
No Fine-Tuning Required
Expert behavior is encoded in text, not model weights. Any sufficiently capable LLM can load a skill and immediately perform at expert level in that domain.
Instant Skill Acquisition
Agents acquire skills at runtime, mid-task. An agent that encounters a security issue can buy the Security Audit Skill and immediately apply expert-level analysis.
Expert Monetization
Domain experts encode their workflows and judgment into SKILL.md files and earn revenue every time an agent applies their methodology. Expert knowledge becomes a recurring asset.
Composable Agent Capabilities
Combine multiple skills in a single agent. A software development agent might stack Code Review + Security Audit + Documentation skills for comprehensive analysis.
Frequently Asked Questions
What is SKILL.md?
SKILL.md is an emerging protocol for defining AI agent behavioral specifications as structured Markdown documents. A SKILL.md file contains an agent's identity, workflow, decision rules, and success criteria — everything needed for an AI agent to adopt a specific expert role. ClawMerchants is the first marketplace to commercially distribute SKILL.md files.
What is an agent skills protocol?
An agent skills protocol is a standardized way for AI agents to acquire and apply behavioral capabilities at runtime. SKILL.md is the leading open format: a structured Markdown document that any LLM can incorporate into its context to adopt a specific expert role, workflow, and decision framework. The skills protocol enables a marketplace model where experts monetize their knowledge by encoding it into reusable behavioral specifications.
How are skills different from data assets?
Data assets return structured JSON — live market feeds, security intel, prices. Skills return text — behavioral protocols, workflows, expert frameworks. Both are accessed through the same x402 payment flow, but skills are consumed differently: an agent incorporates skill text into its context to modify its behavior, rather than processing the response as data.
Can I sell my own skills on ClawMerchants?
Yes. Register as a provider, create an asset of type skill, and upload your SKILL.md file or reference your skill endpoint. Set your price per access. ClawMerchants collects payment and forwards 95% to your registered wallet on Base L2 in USDC.
Are skills version-tracked?
Yes — ClawMerchants includes a skill version tracker that monitors updates to skill definitions. When you update your skill, the version number increments and agents can check for newer versions. This ensures agents always have access to the latest methodologies from skill authors.
Explore More
Agent Data Marketplace
Live data feeds for AI agents — DeFi yields, token anomalies, security intel, and market data. Pay per call with USDC micropayments.
x402 Micropayment API
How the x402 HTTP payment protocol works — the standard that powers agent-native commerce across data, skills, and tools.
Quickstart Tutorial
Get your first agent call working in 10 minutes. TypeScript code, copy-paste ready, covers the complete x402 payment flow.
Agent Skills Marketplace Guide
Deep dive into SKILL.md protocols, 3-stage disclosure, and how to buy and sell agent skills on ClawMerchants.