跳到正文
cccskills
免费在 GitHub 公开

genlayer-intelligent-contracts

Design a GenLayer review workflow for agents acting on a blockchain. Define evaluation criteria, evidence formats, decision outputs, and the connection needed to turn review decisions into restrictions or revocation of delegated authority.

中文简介正在准备中。当前显示英语简介。

原文查看原文说明

Internet Court adapter for GenLayer Intelligent Contract supervision. Use to specify agent-performance rubrics, evidence schemas, decision outputs, and ERC-7710 connector expectations, while delegating actual GenLayer contract writing, linting, testing, deployment, and CLI interaction to the official GenLayer skills at https://skills.genlayer.com/.

实测评分 未实测

查看安装方法

适用场景

  • Define review criteria for delegated agents
  • Structure action logs, receipts, and spending evidence
  • Design decisions to continue, warn, constrain, or revoke
  • Plan how ERC-7710 decisions reach a revocation controller

包含的文件(2)

  • SKILL.md6.9 KB
  • references/agent-supervisor-contract.md2.1 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

GenLayer Intelligent Contracts

Use this adapter when GenLayer is the judgment layer for agent behavior. It should define what the supervisor contract must review and emit: the user's mandate, the agent's action log, transaction receipts, rationales, artifacts, outputs, and the decision about whether authority should continue, be warned, be constrained, or be revoked.

This skill is not the Intelligent Oracle skill. Intelligent Oracle is for binary public web evidence markets. Use this skill for open-ended review of performance, relevance, quality, safety, and mandate compliance.

This skill is also not the GenLayer deployment skill. For writing, deploying, testing, and interacting with GenLayer Intelligent Contracts, use the official GenLayer skills at https://skills.genlayer.com/ or the GenLayer docs. If those skills are available in the environment, load them for implementation. If not, tell the user to install/enable them and keep this skill focused on Internet Court interfaces.

Core Model

Mandate + rubric + evidence
  -> Agent executes transactions
  -> Receipts and outputs are submitted
  -> GenLayer Intelligent Contract reviews behavior
  -> Decision: continue, warn, constrain, revoke, or escalate
  -> Internet Court / revocation controller executes the workflow result

GenLayer can judge qualitative criteria. It does not automatically revoke an EVM delegation unless the system includes a concrete cross-chain message, relayer, controller, or account module that consumes the GenLayer decision.

Workflow

  1. Extract the review target:
    • Objective, allowed authority, prohibited behavior, expected outputs, review cadence, evidence sources, and revocation triggers.
  2. Draft a rubric:
    • Include relevance, mandate compliance, transaction safety, budget discipline, artifact quality, evidence completeness, and user-value criteria.
  3. Define evidence schema:
    • Use structured action receipts, transaction hashes, deployed addresses, agent rationales, output artifacts, costs, and prior warnings.
  4. Define decisions:
    • continue, warn, constrain, revoke, and optionally escalate.
  5. Define effect path:
    • For ERC-7710, load ../genlayer-erc7710-connector/SKILL.md to specify how a revoke or constrain decision reaches the EVM revocation controller or delegation manager.
  6. Add auditability:
    • Store the reviewed action ids, score, reasoning summary, decision timestamp, and next review requirements.

Review Input

Use this as the default review request shape:

type AgentPerformanceReviewInput = {
  mandateId: string;
  objective: string;
  authoritySummary: string;
  prohibitedActions: string[];
  rubric: string[];
  reviewWindow: {
    startsAt: string;
    endsAt: string;
  };
  actionReceipts: AgentActionReceipt[];
  outputs: Array<{
    artifactId: string;
    uri: string;
    summary: string;
  }>;
  costs: {
    txCount: number;
    nativeGasSpent: string;
    tokenSpent?: string;
  };
  priorDecisions: AgentReviewDecision[];
};

For wallet-spend supervision, include an on-chain spend snapshot in each review window:

type WalletSpendSnapshot = {
  mandateId: string;
  permissionHash: `0x${string}`;
  sourceChainId: 84532;
  spendReporter: `0x${string}`;
  actionType: `0x${string}`;
  policyHash: `0x${string}`;
  resourceHash: `0x${string}`;
  delegator: `0x${string}`;
  delegate: `0x${string}`;
  asset: `0x${string}`;
  payTo: `0x${string}`;
  spent: bigint;
  maxTotalSpend: bigint;
  maxPerRequest: bigint;
  requestCount: bigint;
  validAfter: number;
  validUntil: number;
  controllerNonce: bigint;
  reviewWindowId: `0x${string}`;
  evidenceBundleHash: `0x${string}`;
  observedAt: number;
};

On Base Sepolia demos, prefer a reporter contract that reads spend state directly from a permission/accounting manager and sends this snapshot through the GenLayer Studio bridge boilerplate. x402 is one source; contract-call, deployment, compute-credit, or subscription managers can expose the same shape.

Decision Output

type AgentReviewDecision = {
  mandateId: string;
  decision: "continue" | "warn" | "constrain" | "revoke" | "escalate";
  score: number;
  reasoning: string;
  violations: string[];
  constraints?: {
    maxTxsPerPeriod?: number;
    maxGasPerTx?: string;
    requireHumanApprovalFor?: string[];
    disabledActions?: string[];
  };
  reviewedActionIds: string[];
  issuedAt: string;
};

Use constrain when the agent is useful but overreaches. Use revoke when the agent violates scope, creates unsafe deployments, repeatedly wastes budget, hides evidence, or produces outputs that do not meaningfully pursue the mandate.

Rubric Defaults

For delegated on-chain agents, include:

  • Mandate relevance: actions are connected to the user's objective.
  • Scope compliance: actions stay within allowed chains, contracts, selectors, factories, and action types.
  • Safety: no ownership loss, unlimited approvals, suspicious external calls, or avoidable irreversible effects.
  • Budget discipline: gas/value/paid-tool spend is proportional to the work.
  • Evidence quality: every action has a receipt, rationale, and artifact when applicable.
  • Result quality: deployed contracts or outputs are usable and aligned with the requested workflow.
  • Transparency: the agent explains why it took each meaningful action.

Revocation Integration

When the decision can affect ERC-7710 authority, always specify:

type RevocationEffect = {
  sourceChain: "genlayer";
  targetChain: "base" | "ethereum" | string;
  revocationController: `0x${string}`;
  mandateId: string;
  permissionHash: `0x${string}`;
  requiredDecision: "revoke" | "constrain";
  relayer: string;
  finalityAssumption: string;
  fallback: "owner_emergency_revoke" | "absolute_expiry" | "human_review";
};

Do not claim revocation is trustless unless the bridge, relayer, and verification path are defined.

The connector skill owns the exact payload, relay/proof mode, EVM controller interface, and where the ERC-7710 redemption path checks the controller.

Output Checklist

When designing a GenLayer Intelligent Contract review flow, return:

  1. Plain-English review purpose.
  2. Review input schema.
  3. Rubric.
  4. Decision output schema.
  5. Revocation/constraining effect path, with connector skill handoff if ERC-7710 is involved.
  6. Evidence submission process.
  7. Happy path.
  8. Failure and appeal paths.
  9. Tests for continue, warn, constrain, revoke, missing evidence, and bad relayer behavior.

References

  • references/agent-supervisor-contract.md for a reusable supervisor pattern.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Build chat, image generation, and speech transcription features on 0G Compute Network, a distributed GPU marketplace. This skill guides SDK integration, model fine-tuning, response processing for billing, and account funding through code patterns and CLI commands.

中文简介正在准备中。当前显示英语简介。

  • Integrating chat and image generation into an application
  • Preparing model fine-tuning jobs
  • Checking balances and funding provider accounts
internet-court/internet-court-skill70062026年8月20日 更新

Build servers and clients for A2A, an open protocol that lets AI agents communicate across frameworks. Publish agent capabilities, exchange tasks and results, stream progress, and add notifications for workflows involving multiple agents.

中文简介正在准备中。当前显示英语简介。

  • Connect agents built with different frameworks
  • Publish agent capabilities through an Agent Card
  • Build a research, writing, and review workflow
internet-court/internet-court-skill70062026年8月20日 更新

Give an AI agent a wallet with administrator-defined transaction limits. The skill covers encrypted key storage, balance checks, and sending funds through code-enforced rules for destinations, amounts, transaction frequency, and emergency shutdown.

中文简介正在准备中。当前显示英语简介。

  • Give an agent a wallet with transaction limits
  • Restrict transfers to approved chains and addresses
  • Check balances and active rules before sending
internet-court/internet-court-skill70062026年8月20日 更新

Manage a crypto wallet with the awal CLI, including sign-in, balance checks, transfers, swaps, and funding. Also find and pay for APIs through x402, a payment protocol, set up paid endpoints, and query blockchain data on Base.

中文简介正在准备中。当前显示英语简介。

  • Check wallet balances across supported networks
  • Send tokens to an address or ENS name
  • Swap tokens or add wallet funds
internet-court/internet-court-skill70062026年8月20日 更新

Write TypeScript, Rust, or Python code for Alkahest escrow contracts, which hold assets until conditions are met. Covers creating escrows, submitting fulfillment, collecting payments, and building custom contracts that decide whether requirements have been satisfied.

中文简介正在准备中。当前显示英语简介。

  • Integrate Alkahest escrow into an application
  • Build a bot that creates escrows and submits fulfillment
  • Implement payment collection and fulfillment monitoring
internet-court/internet-court-skill70062026年8月20日 更新

Use the Alkahest CLI to manage escrow, where assets are held until agreed conditions are met. It covers buyer deposits, seller fulfillment and collection, third-party decisions, token swaps, and reclaiming expired, unfulfilled deposits.

中文简介正在准备中。当前显示英语简介。

  • Creating escrow with release conditions
  • Submitting work and collecting escrowed assets
  • Approving or rejecting fulfillment as an oracle
internet-court/internet-court-skill70062026年8月20日 更新

查看 internet-court 的全部技能

报告此技能的问题