图谱

SKILLAI 工具
AI 工具 未标注权限 许可证不明确
安装这个技能

把区块链数据问题路由到对的Graph Protocol服务,返实时数据。Route any blockchain data question to

一句话选型结论

按资料字段整理

图谱是AI 工具类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类AI 工具 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install zcwl/graph-advocate
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

AI 工具
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方zcwl仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 enterprise
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验未标注上游最后核验时间 · 数据更新 1791505121
正文体量14.8 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
组合多个 AI 能力

把写作、检索、总结串成一条流程。

搭工作流

把固定流程做成可重复调用的步骤。

做提示词与评测

整理提示词并对比输出效果。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在AI 工具场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
图谱当前 未标注权限 未标注 未标注 未标注 有
执行计划 只读 GitHub 仓库 151.5 万 Codex、Claude、Cursor 有
Playwright MCP 测试 浏览器控制 ClawHub 市场 136.5 万 OpenClaw 有
完成前的验证 未标注权限 GitHub 仓库 131.9 万 Codex、Claude、Cursor 有
子 Agent 驱动开发 只读 GitHub 仓库 131.2 万 Codex、Claude、Cursor 有
前端设计 执行命令 GitHub 仓库 125.6 万 Codex、Claude、Cursor 有

SKILL.md 全文

14.8 KB
模型实际读到的说明书原文,来自技能仓库

Graph Advocate

Ask any blockchain data question in plain English. Get back live data — not just a recommendation.

Routing

Match the user's intent to the right service. Load only the reference you need.

Intent Service Reference Use for
Token balances, holders, swaps, NFTs token-api token-api.md Wallet data across EVM, Solana, TON
Polymarket markets, OHLCV, P&L, positions token-api token-api.md REST endpoints under /v1/polymarket/* — no npm install
Find a subgraph for a protocol subgraph-registry subgraph-registry.md Search 15,500+ subgraphs by protocol/chain
Aave lending data graph-aave-mcp aave.md 40 tools — V2/V3/V4, liquidations, rates
Polymarket advanced (orderbook, disputes, trader winrate/drawdown) graph-polymarket-mcp polymarket.md 31 tools — live CLOB, UMA resolution, subgraph-specific P&L stats
Hyperliquid perps — trader skill, vaults, liquidations token-api + /hyperliquid/* hyperliquid.md Raw markets/users/vaults via token-api; derived skill scores via GA's paid /hyperliquid/* endpoints
Cross-protocol lending graph-lending-mcp — Messari standardized — 40+ protocols on 15 chains
Limitless prediction markets graph-limitless-mcp — Markets on Base
Cross-venue prediction-market spread (Polymarket ↔ Limitless) /predmarket/spread (paid) — Same-topic markets paired across venues, per-pair yes-mid spread (bps), arbitrage direction. JOIN single-venue passthroughs can't return. $0.05 USDC.
Cross-source prediction-market spread (Kalshi ↔ Polymarket) /kalshi-polymarket/spread (paid) — Same-topic markets across Kalshi + Polymarket with mid-spread and arbitrage direction. $0.05 USDC.
Agent reputation score (0-100) /agent/score (paid) — Composite from ERC-8004 registration + USDC settlement velocity + on-chain feedback registry, aggregated across all of owner's agents. Hard 8004 gate filters burns/CEX. Buyer-agent's "should I trust this peer" call. Response includes counterparty-ref-v1 conformant SHA-256 + preimage + canonical JCS string so a verifier can re-derive byte-exact without callback. $0.02 USDC.
Kalshi consensus trend /kalshi/consensus-trend (paid) — Slope/acceleration/volatility band from Kalshi forecast_history. $0.05 USDC.
Kalshi sports live-edge /kalshi/sports-live-edge (paid) — Play-by-play momentum vs market candlesticks; flags latency-arb windows. $0.05 USDC.
Predict.fun prediction markets predictfun-mcp — BNB Chain prediction markets
x402 payment analytics — NL question /ask (paid) — Natural-language Q&A over 132M settlements + daily_stats (May 2025 → Jun 2026). Sonnet + DuckDB. $0.05 USDC.
x402 address lookup — onchain receipts /onchain-x402/address (paid) — Decentralized lookup via x402 Base subgraph: lifetime stats by role (payer/recipient), recent payments, facilitator, indexed_through_block. $0.01 USDC.
x402 payment analytics (legacy reference) x402-analytics x402.md Payment volume, facilitators, daily stats on Base
Raw block data, streaming substreams — Traces, logs, custom transformations
Agent discovery (ERC-8004) 8004scan — Find AI agents by capability
MCP server auth mcp8004 — ERC-8004 identity verification

Polymarket routing rule: Prefer token-api for common queries (markets, OHLCV, activity, user positions, P&L, platform stats). Only route to graph-polymarket-mcp for advanced queries: live orderbook depth, live spreads, disputed markets, UMA resolution, trader winrate/drawdown/profit factor, CTF splits/merges/redemptions.

If the request spans two services, use both and combine results.

Quick Examples

"Top 10 USDC holders on Ethereum"           → token-api
"Best subgraph for Uniswap V3 on Arbitrum?" → subgraph-registry
"Aave V3 liquidations above $50K"           → graph-aave-mcp
"Hottest Polymarket markets"                → token-api (/v1/polymarket/markets)
"Polymarket OHLCV for Bitcoin market"       → token-api (/v1/polymarket/markets/ohlc)
"Polymarket trader P&L for 0x..."           → token-api (/v1/polymarket/users/positions)
"Polymarket live orderbook depth"           → graph-polymarket-mcp (advanced)
"Polymarket trader winrate/drawdown"        → graph-polymarket-mcp (subgraph P&L stats)
"Score Hyperliquid trader 0x..."            → /hyperliquid/score (paid)
"Hyperliquid top traders for HYPE"          → /hyperliquid/screen (paid)
"Evaluate Hyperliquid vault 0x..."          → /hyperliquid/vault (paid)
"Compare Aave vs Compound TVL"              → graph-lending-mcp
"x402 payment volume on Base today"         → x402-analytics
"Top 10 x402 recipients in the last 30 days" → /ask (paid, NL→SQL)
"When did x402 volume on Base inflect?"     → /ask (paid, NL→SQL)
"Has 0x0FF5A6… ever been paid via x402?"     → /onchain-x402/address (paid, decentralized)
"Polymarket vs Limitless spread on 'trump'"  → /predmarket/spread (paid, cross-venue JOIN)
"Kalshi vs Polymarket fed-rate arbitrage"    → /kalshi-polymarket/spread (paid)
"Find agents that do trading"               → 8004scan

How It Works

  1. Agent sends plain-English question
  2. Graph Advocate identifies the best service
  3. Searches the subgraph registry (15,500+ subgraphs with query hints)
  4. Executes the query and returns live data in the response
  5. Includes get_started link for agents to get their own free API key

Response Format

{
  "recommendation": "subgraph-registry",
  "reason": "why this service fits",
  "confidence": "high",
  "query_ready": { "tool": "...", "args": {...} },
  "execution_result": { "source": "subgraph-gateway", "data": {...} },
  "get_started": "Free API key: https://thegraph.com/studio/",
  "cache_for_seconds": 86400
}

Endpoints

Method URL Purpose
POST https://graphadvocate.com/ A2A JSON-RPC 2.0
POST https://graphadvocate.com/chat Simple HTTP chat
GET https://graphadvocate.com/.well-known/agent-card.json Agent card
GET https://graphadvocate.com/agents/capabilities.json Machine-readable capability list
GET https://graphadvocate.com/mcp/catalog List of installable MCP servers
GET https://graphadvocate.com/llms.txt LLM-friendly discovery file
GET https://graphadvocate.com/quota?sender=0x... Free-tier quota remaining today (no charge)
GET https://graphadvocate.com/dashboard Live monitoring
POST https://graphadvocate.com/feedback Agent feedback

Default mode: no wallet, free tier

This skill is fully functional with no wallet attached. The default routing —
plain-English question in, JSON recommendation + live data out — runs against the
free tier and never asks for funds. Agents that don't expose an x402-enabled
runtime simply never see a payment challenge.

If you're evaluating Graph Advocate for the first time, start without a wallet,
use the free tier, and decide later whether the paid endpoints are worth opting
into.

Optional paid mode (opt-in)

Some endpoints settle in USDC on Base via the x402 protocol. Paid mode is
opt-in: it only activates when your agent runtime is configured to accept
x402 payment challenges. Without that configuration, paid endpoints return
402 Payment Required and the call stops there — no funds move.

Pricing

  • /route — 3 free queries/sender/day, then $0.01 USDC per call (Base mainnet)
  • /polymarket/* — paid from call 1 ($0.01 - $0.05 per call)
  • /hyperliquid/* — paid from call 1 ($0.02 - $0.10 per call)
  • /predmarket/spread — paid from call 1 ($0.05 USDC) — Polymarket ↔ Limitless cross-venue spread on a topic. POST {topic, limit?} returns per-pair yes-mid spread (bps) and arbitrage direction. JOIN that single-venue passthroughs structurally can't return.
  • /kalshi-polymarket/spread — paid from call 1 ($0.05 USDC) — Kalshi ↔ Polymarket cross-source spread on a topic.
  • /kalshi/consensus-trend, /kalshi/sports-live-edge — paid from call 1 ($0.05 USDC each).
  • /ask — paid from call 1 ($0.05 USDC) — natural-language Q&A over the x402 Base settlements warehouse. 132M+ payments + pre-aggregated daily_stats (388 days). Returns {answer, sql_trace, model, upstream_ms} so callers can verify the data path.
  • /onchain-x402/address — paid from call 1 ($0.01 USDC) — decentralized address lookup against the x402 Base subgraph on The Graph Network. POST {address} returns lifetime stats (payer + recipient roles), recent 10 payments in each direction, facilitator metadata, and indexed_through_block for freshness.

402 challenge — preview before paying

Every paid endpoint's 402 response body includes an output_example field with a sample of the payload you'll receive after paying. Inspect this before signing — if the shape doesn't fit your use case, bail without spending.

Per-call approval (recommended)

Every paid call is preceded by an HTTP 402 challenge that names the price,
the recipient address, and the settlement network:

HTTP/1.1 402 Payment Required
X-Payment: { "amount": "0.01", "currency": "USDC", "network": "base",
             "recipient": "0x0FF5A6ec…7C86", "challenge": "<nonce>" }

The 402 challenge is the per-call approval surface. Configure your runtime
to surface 402 challenges to the user
before any payment is signed — most
x402 clients accept an interactive-approval flag (e.g. confirmBeforePay).
With per-call approval enabled, you see the exact price and recipient on every
paid request and decide yes/no per call. This is the recommended posture for
any wallet that holds more than a trivial amount.

If you instead configure your runtime to accept 402 challenges automatically,
paid calls settle without an interactive prompt. That is a runtime-level choice,
not a property of this skill — and if you choose it, the spend controls below
become load-bearing.

Free-quota visibility (no-charge)

Before any paid call, check today's remaining free-tier quota:

GET https://graphadvocate.com/quota?sender=0x<your-agent-address>

Returns:

{
  "sender": "0x…",
  "free_quota_daily": 3,
  "used_today": 1,
  "remaining_today": 2,
  "free_tier_exhausted": false,
  "next_call_paid": false,
  "price_usdc_per_paid_call": 0.01
}

/quota is a no-charge metadata route. Use it to:

  1. Display "N free queries remaining today" in your UI before the agent runs.
  2. Halt autonomous loops when remaining_today hits 0 instead of accepting
    the 402 challenge.
  3. Audit per-sender spend by polling daily.

Required spend controls before autonomous use

  1. Dedicated low-balance wallet. Fund only what you're willing to spend
    in a session (e.g. $5 USDC). Never connect a wallet that holds your treasury.
  2. Spend caps in your agent runtime. Most x402 clients accept
    maxAmountPerCall and maxTotalSpend parameters. If yours does not, wrap
    calls to this skill in a counter that breaks after N invocations.
  3. Stop conditions. Add at least one of:
    - Hard cap on call count per task (e.g. max_calls_per_run = 5)
    - Time bound on the task (e.g. abort after 5 minutes)
    - Cost ceiling check before each call (read wallet USDC balance; halt if below threshold)
  4. First-run probe mode. Treat the first execution as a probe — log every
    call, inspect spend, then enable autonomous mode only after confirming the
    per-task cost matches expectations.

Receipts and no-charge surface

  • Every paid call returns an x-payment-response header with the on-chain
    settlement reference. Log it. If a call doesn't return that header but charged
    your wallet, file an issue — the contract is "no settlement, no charge."
  • No-charge endpoints never trigger payment: /.well-known/agent-card.json,
    /agents/capabilities.json, /mcp/catalog, /llms.txt, /dashboard,
    /chat, /quota, GET /.

Payments are received by Ampersend smart account 0x0FF5A6ecef783BBA35463ec2F8403B9B5e9e7C86.

External Endpoints

Endpoint Data sent Purpose
graphadvocate.com Your plain-English query Routes to the right Graph service
gateway.thegraph.com/api/ GraphQL queries Executes subgraph queries for live data
api.pinax.network/ REST requests Fetches token/NFT/swap data
api.studio.thegraph.com GraphQL queries x402 payment analytics

Security & Privacy

  • Instruction-only skill — no code is downloaded or executed on your machine
  • No credentials required — Graph Advocate does not need API keys from you
  • No local file access — reads nothing from your filesystem
  • Stateless — no session data persists between requests

Identity

  • ERC-8004: Agent #734 (Arbitrum), #41,034 (Base)
  • ENS: graphadvocate.eth
  • Ampersend: app.ampersend.ai/discover/agent/8453:41034

Trust Statement

By using this skill, your plain-English data queries are sent to graphadvocate.com (hosted on Railway, operated by @paulieb14). The service returns structured JSON with live data. Queries may include wallet addresses and protocol/trading intent — do not send sensitive private context (private keys, seed phrases, internal strategy details) and only install if you trust this endpoint operator.

Paid mode is opt-in. This skill never requests credentials from you. Paid
endpoints only settle when your agent runtime is configured to accept x402
payment challenges. Configure interactive approval if you want a per-call
y/N prompt, and always use a low-balance wallet — see the "Optional paid mode"
section above for the recommended posture and spend controls.

Links

  • GitHub: https://github.com/PaulieB14/graph-advocate
  • The Graph: https://thegraph.com
  • Subgraph Studio: https://thegraph.com/studio

常见问题

5 问
答案由本站依据条目字段整理
图谱怎么安装?

在支持 Skills 的客户端里执行:skillhub install zcwl/graph-advocate。安装方式为未标注。装完刷新技能列表或重开客户端。

图谱支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

图谱权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

图谱收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

图谱用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 未标注。