Research hub

AI skill marketplace guides for buyers, sellers, and research agents

This guide hub explains the Sealed thesis in public, metadata-only language: an AI-agent skill marketplace can be discoverable and useful without exposing the protected skill body. The practical answer is server-side execution, current safety screening, per-call pricing, and agent-readable discovery files where skill bodies never ship.

Browse screened skills Read metadata catalog Read agent discovery file

Guide cluster

Seven high-intent pages cover the core marketplace questions: what exists, why sealed execution matters, how sellers earn, and what agents can safely discover.

  1. AI agent skill marketplace: what buyers and sellers needThe marketplace trust model: screened skills, metadata-only discovery, and server-side execution.marketplace trust
  2. Enterprise AI skill marketplace: safe agent skills at workHow enterprise teams evaluate skill marketplaces for auditability, policy controls, and source protection.enterprise safety
  3. How to sell prompts without them getting copiedWhy downloadable prompt markets leak, and how sellers can earn without handing over the method.seller IP
  4. Server-side skill execution, explainedThe core containment model: buyers send inputs, Sealed runs the skill, buyers receive outputs only.architecture
  5. Per-call pricing for AI skillsWhy pay-per-use works when the marketplace sells each useful run instead of the underlying file.economics
  6. What you can and can't protect when you sell an AI skillThe doing-vs-dispensing test for deciding which skills are structurally protectable.seller strategy
  7. Skills vs MCP tools vs promptsHow the three capability formats differ, and why a hosted skill marketplace fills the missing middle.market map

Agent discovery without source leakage

Agents and crawlers get useful public context. Protected skill source remains outside every public route.

Metadata-only catalog

/skills.json exposes public skill names, descriptions, input schemas, prices, safety status, and links. It never includes a skill body.

Canonical marketplace

/app is the primary product surface for browsing screened skills, inspecting trust signals, and starting a safe run path.

Research-agent orientation

/llms.txt and /sitemap.xml point agents toward public pages while stating that skill bodies never ship.

Machine-readable paths: /skills.json, /llms.txt, /sitemap.xml, /app, and per-skill pages at /skills/{id}/page.

Trust boundary: public discovery describes capabilities; skill source stays sealed and only enters the server-side execution path.