SkillMint

SkillMint is a verifiable AI skill execution protocol built on 0G for the agent economy.

It solves a core trust and monetization gap in open AI-agent ecosystems: today, when agents call third-party skills, there is often no cryptographic

proof of what actually ran, no durable execution receipt, and no reliable settlement rail between skill creators and consumers.

SkillMint turns AI skills into on-chain assets and enforces verifiable execution + programmable settlement end-to-end.

A creator publishes a skill with on-chain pricing, and the skill is represented as an NFT in the Skill Registry.

A user or agent executes the skill through escrowed payment flow.

Execution is processed through 0G Compute with TEE attestation checks, producing a verifiable result path.

A canonical receipt is serialized, uploaded to 0G Storage, and its root hash is anchored on-chain.

After confirmation, escrow auto-settles funds with protocol split and payout to the skill owner.

This architecture gives three guarantees:

1. Execution integrity: inference path is attested and independently checkable.

2. Receipt durability: immutable execution records are persisted in 0G Storage and linked on-chain.

3. Trustless settlement: payments are released through audited contract logic, not manual backend accounting.

SkillMint is designed for both human and agent-native usage.

It supports multi-asset pricing/settlement rails (0G, W0G, USDC.E stablecoin) and includes x402-compatible payment flow for autonomous agent

transactions.

It also provides a TypeScript SDK for discovery, execution, verification, and integration into agent frameworks.

0G components integrated:

- 0G Chain: skill registry, escrow, settlement, on-chain execution anchoring.

- 0G Compute: TEE-attested inference workflow.

- 0G Storage: encrypted skill artifacts and execution receipt persistence.

As of submission, SkillMint is live on 0G mainnet with deployed contracts, live skill executions, and on-chain settlements, demonstrating a working

publish → execute → verify → settle

SkillMint was built from scratch during the hackathon and iterated continuously .

Development started with the initial protocol commit and quickly moved into core execution trust features, including real input passthrough and

end-to-end prompt encryption.

Over the next phases, we shipped and hardened the full stack: oracle execution reliability, verification improvements, and edge-case handling; x402

payment infrastructure; SDK ergonomics; and deployed backend services (oracle + facilitator) on EC2. We also expanded functionality beyond basic prompt

skills by adding native agent-skill bundle support across the stack.

As the build matured, we migrated to 0G Aristotle mainnet and switched product defaults to mainnet, then introduced V3 contract architecture with dual

pricing and multi-token settlement. This included SkillRegistryV3, SkillEscrowV3, FACILITATOR_ROLE support, and stablecoin rails (USDC.E) alongside

0G/W0G. We then wired V3 addresses into deployment/config surfaces and finalized UI/verification updates for token-aware settlement display and

execution consistency.

In short, hackathon progress was not limited to a prototype UI: we delivered an end-to-end, production-style protocol flow and iterated it into a live

mainnet-ready system across contracts, backend services, SDK, and frontend.

Bootstrapped pre-seed. Built solo during the hackathon, no external funding.Open to grant / accelerator conversations post-submission.