What is an AI agent skill?

Plain English first

If you buy software for yourself, you usually get a license, a download, or a login. If you build AI agents, you need something similar: a clear package the agent can fetch, evaluate, and run without a human clicking through a SaaS dashboard every time.

LVL skills are those packages. Each skill has a stable id, a price in USD, a free outline (and often a sample), and a sealed implementation that is only returned after a verified on-chain payment. Humans and agents use the same rails.

Capability packs vs prompt packs vs plugins

Prompt packs

A prompt pack is usually a collection of text instructions. Useful, cheap to ship, easy to copy. Weaknesses: no versioned deliverable, no integrity check, no payment proof, and no standard install surface. Anyone can paste the same text into a chat window and call it done.

Plugins and SaaS tools

Plugins and hosted APIs are powerful, but they often require long-lived API keys, monthly seats, and vendor dashboards. That model fights autonomous agents. An agent that must store your OpenAI key and a dozen other secrets is an ops burden and a security surface.

LVL capability packs (skills)

A skill on LVL is meant to be a sealed capability unit: documentation plus implementation files you only get after payment. The free side is honest evaluation. The paid side is the sealed pack. Payment is HTTP 402 style commerce on Base USDC (the x402 pattern), not a card form buried in a checkout SPA.

• Free: outline JSON, sample markdown, listing page, machine catalog entry
• Paid: sealed_pack.files after verified Base USDC transfer
• Proof: public ledger at /api/proof for confirmed unlocks only
• Re-redeem: same payment can unlock again (idempotent path)

What you see before you pay

Every skill should be evaluable without a wallet. Open the listing or the free outline. Read the summary, when-to-use notes, and sample. If the outline is thin or templated, treat that as a quality signal and walk away. LVL is actively pushing unique outlines; free evaluation exists so you can refuse filler.

What you get after you pay

After the server verifies your transaction on Base, the unlock response includes sealed pack files (for example SKILL.md and implementation assets). That is the product. You can re-request with the same proof so a flaky download does not strand you.

Categories on the shelf

Skills are organized into practical lanes such as Research, Memory, Finance, Coding, Networking, Neurotech, and Browser. Categories help humans browse and help agents filter catalog search. They are not certifications. They are shelf labels.

What a skill is not

• Not a claim of third-party audit badges unless a listing explicitly documents evidence
• Not a hosted always-on SaaS seat (unless a separate product says so)
• Not invented sales volume; check /api/proof for confirmed unlocks
• Not the old merch brand sometimes still cached in search results (see /status.json)

How to go deeper

If you are buying by hand, read the MetaMask buyer guide next. If you are wiring an agent, read the autonomous HTTP purchase guide and the discovery guide for catalog.json and llms.txt. Both paths end at the same payment truth: amount, payTo, asset, network, then proof.

A day in the life: human buyer

You open the marketplace, filter by Networking, and open a listing. The page shows a green USDC price, a free sample link, and a buy button. You read the sample. If it is specific, you connect MetaMask on Base, pay once, and download files. You never create a SaaS seat. You never hand the seller a long-lived API key just to receive the pack.

That flow is the product promise. Skills that break it (no free eval, unclear files, bait pricing) should be treated as defects, not as normal marketplace chaos you have to accept.

A day in the life: agent buyer

Your agent wakes with a budget and a goal: "acquire a post-meeting debrief skill under $5." It calls GET /api/shop, searches catalog JSON, fetches two outlines, scores them with your rubric, issues GET /api/pay for the winner, sees HTTP 402, transfers USDC, POSTs the hash, and writes sealed files into its workspace. No Puppeteer required for the happy path.

If the agent cannot complete that loop, the failure is usually wallet policy, wrong network, or missing gas, not missing HTML selectors.

Versioning and identity

Each skill has a stable string id such as seller-agent or agent-x402-first-buy. That id appears in catalog.json, pay challenges, proof rows, and file paths. Do not rename ids casually. Titles and seo_title fields can change for humans; ids are the contract for agents.

When a skill updates, buyers who already unlocked may need a migration note. LVL documentation encourages clear version communication. If a listing is silent on versions, assume you are buying the current sealed snapshot at pay time.

Composability

Serious agent systems compose many skills: research then write then file. Bundles on LVL package multiple skills with an execution order and a single payment. Even without bundles, your orchestrator can buy units independently and wire them. The skill abstraction is the unit of commerce; your graph is the unit of work.

Quality signals that are free

Before you spend, look for: unique outline steps, concrete inputs and outputs, failure modes, price coherence with tier tags, and machine links that 200 OK. After small spends, look at /api/proof for whether anyone else confirmed unlocks. Zero unlocks is not automatic fail for a new skill, but it is information.

Also read status.json if search engines still show outdated merch branding. The live product is an agent skill market.

Common misconceptions

• "Paying unlocks a hosted API forever." Usually you receive files. Hosted runtime, if any, will be spelled out explicitly.
• "Outline equals full pack." No. Outline is evaluation. Pack is paid.
• "Agents must scrape the storefront." No. Use catalog and shop JSON.
• "All skills are first-party curated." Open market listings can differ; know which feed you trust.

Checklist before first purchase

1. Confirm you are on lvlltd.com (or a trusted mirror you control).
2. Open free outline for the skill id you intend to pay.
3. Confirm Base network and USDC asset expectations.
4. Start with the $0.05 canary if you have never unlocked before.
5. Save txHash and files; know re-redeem exists.

When that checklist is boring, you are ready for larger packs. Boring is good in commerce rails.

Field notes from live rails

Everything in this guide is meant to match production behavior on lvlltd.com: free outlines at /skills/{id}/outline.json, challenges at /api/pay, settlement in USDC on Base chain ID 8453, unlock proofs that can be checked on BaseScan, and a public ledger at /api/proof that refuses to invent volume. If a third-party article disagrees with those primary surfaces, trust the primary surfaces.

When you teach teammates, walk them through one canary unlock before any architecture debate. Shared scars on a $0.05 purchase create better systems design conversations than slides alone. Keep notes on what confused people; those notes should become checklist items in your internal runbook and, if you sell skills, clearer outline text for the next buyer.

Finally, remember the integrity line that LVL repeats for a reason: understanding is free; sealed capability is optional and paid. Education pages like this one should make you more demanding, not more impulsive. Read outlines. Allowlist addresses. Prefer small spends first. Verify. Then scale.

Related live checks

After reading, run at least two live checks: open the free outline for a skill you care about, and open /api/proof. Optionally issue a pay challenge with curl and confirm you see HTTP 402 with amount and payTo fields. Those three actions connect this article to production reality in under two minutes.

If you maintain internal wiki mirrors of these guides, include the canonical URLs on lvlltd.com so agents and humans can detect drift. Canonicals and llms.txt exist so education stays discoverable next to commerce, not buried in a slide deck.