# Structure Agent Commerce With Internet Court

Agent commerce creates payment, custody, and dispute risk across many protocols. Internet Court gives Claude, Codex, and Claude Code a router for bounded payments, escrow, evidence, and adjudication.

## Install

```bash
npx skillstore add internet-court/internet-court
```

## Metadata

- Status: approved
- Slug: internet-court-internet-court
- Skillstore revision: r1
- Version status: missing
- Tree hash: 370a15e0720f135e57402e9fbddc3af4723c6f041cc94d225f4e1b42b995ec96
- Author: internet-court
- GitHub username: internet-court
- License: MIT
- Repository: https://github.com/internet-court/internet-court-skill/tree/main/
- Ref: 3f6e026a3363e0954ede7bef0cfe88d4475de137
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: blocked
- Manual install advisory: allowed\_with\_warning
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, external\_commands, filesystem, env\_access, scripts
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/internet-court-internet-court
- Manifest: https://skillstore.pages.dev/api/skills/internet-court-internet-court/manifest

## Capabilities

- Routes agent commerce requests to vendored protocol skills only when the task reaches that layer.
- Guides discovery-first review of counterparties, paid endpoints, advertised prices, and promised deliverables.
- Defines Basic, Guarded, and Adjudicated trust levels for agent deals.
- Connects x402, ERC-7710, GenLayer, escrow, and reputation patterns into one workflow.
- Sets wallet safety boundaries, proof requirements, and user approval expectations before transactions.

## Use Cases

- Assess a Paid Agent Service: Inspect a paid endpoint, identify its payment rail, and list missing trust facts before spending.
- Design a Bounded Payment Flow: Plan an x402 or delegated wallet payment with caps, allowed routes, and approval boundaries.
- Prepare a Dispute Path: Define evidence, review criteria, settlement actions, and reputation updates for contested work.

## Prompt Templates

### Map a Deal

```
Use Internet Court to map this agent-to-agent transaction: [service, counterparty, cost, and deliverable]. Start with discovery only and list missing trust facts.
```

### Choose a Trust Level

```
Given this proposed agent deal: [details], choose Basic, Guarded, or Adjudicated protection. Explain the tradeoff and what must be confirmed first.
```

### Prepare Guarded Payment

```
Design a guarded payment flow for [paid resource]. Include the payment rail, spending cap, allowed resources, wallet approval boundary, and failure handling.
```

### Design Adjudicated Settlement

```
Create an adjudicated workflow for [agent job]. Define evidence, review rubric, decision outcomes, enforcement action, and reputation update requirements.
```

## Limitations

- It does not sign transactions or approve wallet prompts for the user.
- Many routed sub-skills require provider credentials, funded test wallets, or external services.
- Some standards named by the skill do not yet have public bundled skills.
- Root-only installs may fetch sibling skills from the network unless the full repository is installed.

## Best Practices

- Install the full repository when possible so routed sub-skills are local and reviewable.
- Keep the agent in discovery mode until the user grants funds, wallet authority, or explicit permissions.
- Use the lightest trust level that covers the real payment, custody, and delivery risk.

## Anti Patterns

- Do not ask users for private keys, seed phrases, or unrestricted bearer wallet artifacts.
- Do not fetch or execute remote installer scripts without checksum or signature verification.
- Do not claim a payment, delegation, contract, or dispute result exists before receiving proof.

## Security Audit

- Audited at: 2026-07-19T10:16:26.534\+00:00
- Summary: The bundle contains confirmed critical remote installer pipelines, secret-retrieval commands, model-generated eval, plaintext private-key workflows, and silent global plugin installation. Many remaining matches are lexical false positives, but the confirmed supply-chain and credential risks prevent unattended use. Static review was capped at 400/6908 representative findings; omitted static matches are unconfirmed, so automatic publishing stays disabled until manual review.

## Stats

- Views: 0
- Downloads: 0
- Favorites: 1
- Popularity score: 0