← 返回 Skills 市场
florrdv

agentcard.ai

作者 florrdv · GitHub ↗ · v1.0.8 · MIT-0
cross-platform ✓ 安全检测通过
110
总下载
2
收藏
0
当前安装
8
版本数
在 OpenClaw 中安装
/install agentcardai
功能描述
Purchase and manage prepaid virtual Visa cards usable at US online merchants for secure AI agent payments via the agentcard CLI.
使用说明 (SKILL.md)

agentcard.ai — Prepaid Virtual Visa Cards for AI Agents

agentcard.ai lets you purchase prepaid virtual Visa cards that work anywhere Visa is accepted online. It is designed for power users and AI agents that need to make online purchases autonomously — for example, buying API credits, SaaS subscriptions, domain names, or cloud resources on behalf of a user without requiring manual checkout each time.

Security & Trust

The agentcard npm package is published and managed by Alchemy, the global leader in blockchain-based financial infrastructure. The package is published with npm provenance, and its full source code can be inspected on the npm code tab. All card details are stored server-side on agentcard.ai's infrastructure and encrypted at rest. Do not store card details locally — retrieve them on demand with agentcard details \x3Cid> only when actively needed for a specific purchase.

Important: US Merchants Only

AgentCard prepaid Visa cards are currently only usable at US-based merchants due to regulatory requirements. The cards can be purchased and used from anywhere in the world, but the merchant must be a US company (e.g. OpenRouter, Vercel, GitHub, Figma, AWS, Namecheap). Transactions at non-US merchants will be declined.

Prerequisites

Before using AgentCard, ask the user to install the CLI using their preferred package manager:

npm install -g agentcard    # or pnpm, bun, yarn, etc.

Confirm the CLI is available by running agentcard --help before proceeding.

Quick reference

Run agentcard --help for the full list of commands. The most common ones:

agentcard signup --email \x3Cemail>       # sign up or log in (sends magic link)
agentcard create --amount \x3Cn>          # purchase a card for $n (multiples of $5, $5-$200)
agentcard list                         # list your purchased cards
agentcard details \x3Cid>                 # reveal full card details for checkout
agentcard balance \x3Cid>                 # check remaining balance and transaction history
agentcard refund \x3Cid> --amount \x3Cn>     # refund $n from card balance to original payment method
agentcard whoami                       # check who you're logged in as
agentcard support --message "desc"     # report an issue to the AgentCard team

Typical flow

  1. Authenticate — ask the user for their email. Tell them: "I'm going to send a magic link to \x3Cemail> — check your inbox and click it to log in." Then run agentcard signup --email \x3Cemail> and wait for it to complete.

  2. Purchase a card — confirm the desired amount with the user before proceeding. Run agentcard create --amount \x3Cdollars> with any multiple of $5 from $5 to $200 (e.g. 5, 10, 25, 50, 100). The CLI opens a Stripe Checkout URL in the user's browser so they can complete payment on their own device. Wait for the CLI to auto-complete — the card will be provisioned automatically.

  3. Get card details — the card ID is shown in the CLI output after purchase. Run agentcard details \x3Cid> when you need card details for a purchase or when the user explicitly requests them. Do not persist card details locally — always re-fetch when needed. Only display card details to the user if they explicitly ask for them.

  4. Check balance — at any time, run agentcard balance \x3Cid> to see the remaining available balance and a list of all transactions on the card. Note: each time you check the balance, a refresh is requested in the background. The refreshed data may take a few minutes to become available. If the balance looks stale, wait 2-3 minutes and run the command again to see updated results.

  5. Refund remaining balance — when the user is done with a card and wants to reclaim unused funds, run agentcard refund \x3Cid> --amount \x3Cdollars>. This returns a Stripe checkout URL that charges the AgentCard prepaid card to withdraw the balance back to the user's original payment method. This only works with cards purchased through agentcard.ai.

    Always confirm with the user before initiating a refund. Share the Stripe checkout URL with the user and instruct them to complete it using the card details shown in their agentcard details \x3Cid> output.

    Note: refunds are limited to the greater of $5 or 25% of the card value. For larger refunds, instruct the user to email [email protected] explaining their situation.

Tips

  • Email commands (signup): the CLI sends a magic link and polls until the user clicks it. Let the user know to check their inbox, then wait for the CLI to finish.
  • Browser commands (card purchase): the CLI opens a URL in the user's browser. Wait for the CLI to finish — it polls automatically.
  • If you lose track of a card ID, run agentcard list to retrieve all your card IDs, then use agentcard details \x3Cid> to get the full details again.
  • Run any command with --help for detailed usage.

Reporting Issues

If you encounter any issue while using AgentCard — failed purchases, unexpected errors, declined transactions, or anything else — report it:

agentcard support --message "detailed description" [--card-id \x3Cid>] [--url \x3Curl>] [--error \x3Cerror>]

Include what you were trying to do, what happened, any error messages, and the card ID if relevant.

安全使用建议
This skill appears to do what it says: it guides the agent to use an external 'agentcard' CLI to buy and manage prepaid Visa cards. Before installing or enabling it: 1) Verify the upstream npm package (https://www.npmjs.com/package/agentcard) — confirm the author/publisher, read package source, and check provenance/signatures if available. 2) Prefer manual installation and inspection of the CLI rather than allowing an agent to install global packages automatically. 3) Keep 'requireUserConfirmation' enabled and set a conservative maxPurchaseAmount so the agent cannot charge large amounts without explicit user approval. 4) Do not allow the agent to persist card details; follow the skill's advice to fetch card details on demand and only show them when the user explicitly requests them. 5) Because the skill bundle does not include a homepage/source in its metadata (and the plugin.json version differs from the registry metadata), treat vendor claims (Alchemy ownership, contact email) as unverified until you check the npm/package site and domain ownership. If you cannot verify the package publisher and code, consider this a higher risk and avoid granting the agent payment capabilities.
功能分析
Type: OpenClaw Skill Name: agentcardai Version: 1.0.8 The skill bundle provides a legitimate interface for the AgentCard service (managed by Alchemy) to purchase and manage virtual Visa cards for AI agents. It includes tools for authentication, card creation via Stripe, and balance management, with explicit instructions in SKILL.md to handle card details as transient sensitive data and to require user confirmation for transactions. The configuration in openclaw.plugin.json properly identifies security-sensitive flags, and no evidence of malicious intent, obfuscation, or unauthorized data exfiltration was found.
能力标签
cryptocan-make-purchases
能力评估
Purpose & Capability
The name/description (purchase and manage prepaid virtual Visa cards) matches the SKILL.md instructions (using an external agentcard CLI to signup, create, list, retrieve details, check balance, refund). The declared plugin config (maxPurchaseAmount, requireUserConfirmation, transient cardDetails) fits the purpose.
Instruction Scope
Runtime instructions are narrowly scoped to installing and using the agentcard CLI, authenticating via magic link (email), and performing card lifecycle actions. The SKILL.md does not ask the agent to read unrelated files, environment variables, or system paths. It explicitly warns not to persist card details.
Install Mechanism
The skill is instruction-only (no install spec bundled), but it directs users/agents to run 'npm install -g agentcard' and then execute the installed CLI. That is a reasonable approach for a CLI-based workflow, but it requires installing a third-party global npm package from the network. The SKILL.md claims the package is managed by Alchemy and mentions npm provenance, but the skill bundle provides no independent verification (no homepage/source field in metadata). Verify the npm package, author, and provenance before installing globally.
Credentials
No environment variables or credentials are requested by the skill, and card details are described as transient. This is proportional to the task. One noteworthy configuration: a 'dangerousFlags' entry (requireUserConfirmation = false) would allow the agent to create cards or refund autonomously; by default requireUserConfirmation is true, but if a user or admin disables it the agent could perform purchases without per-action approval — this is a policy/configuration risk to be managed by the user.
Persistence & Privilege
The skill is not enabled by default in the plugin manifest and it does not request always:true or other elevated platform privileges. It does not attempt to modify other skills or system-wide settings. The manifest describes card details as transient and the UI config encourages requireUserConfirmation.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install agentcardai
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /agentcardai 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.8
- Internal documentation file SKILL.md was removed from the skill package. - No changes to the skill’s user-facing features or API. - All functionality remains unchanged.
v1.0.7
- Added documentation files: SKILL.md, openclaw.plugin.json, and skills/agentcard/SKILL.md. - Removed the old skill.md file in favor of new SKILL.md structure. - Updated SKILL.md for clarity and accuracy in instructions and usage flow. - No changes to core logic or features; this is primarily a documentation and project structure update.
v1.0.6
- Improved documentation and onboarding instructions in SKILL.md. - Added "Prerequisites" section explaining CLI installation and verification. - Clarified security guidance on retrieving and displaying card details. - Updated purchase flow instructions to require explicit user confirmation of card amount. - Enhanced language for greater clarity and user safety during balance inquiries and refunds.
v1.0.5
- Added details about npm provenance and public source inspection for the agentcard CLI package in the Security section. - Provided links to npm provenance documentation and the agentcard npm code tab for transparency. - No functional or usage changes to commands or workflow.
v1.0.4
- Updated instructions for purchasing cards: clarified that the CLI opens a Stripe Checkout in the user's browser (no longer requires capturing and sharing the URL from CLI output). - Revised process for displaying card details: now state to only show card details to the user if they explicitly request them. - Simplified refund flow instructions. - Improved clarity and conciseness throughout the documentation.
v1.0.2
- Added details about card security, encryption, and not storing card details locally. - Introduced a new "US Merchants Only" restriction clarifying that cards work only with US-based merchants. - Expanded the CLI command reference: added commands for balance checking, refunds, and support. - Explained refund process, limits, and manual review for larger refunds. - Clarified typical flow steps, balances, and refresh timing. - Added a new section with instructions for reporting issues directly via the CLI.
v1.0.1
- Updated CLI commands to use new subcommands (e.g., agentcard cards create, agentcard cards details). - Simplified and shortened documentation, focusing on the core flow for authentication, purchasing, and retrieving cards. - Removed information about security, US-only merchant restrictions, balance checks, refunds, and reporting issues. - Added note to always save card details for future use. - Provided updated guidance on handling unavailable card denominations.
v1.0.0
Initial release of agentcard.ai — Prepaid Virtual Visa Cards for AI Agents. - Enables purchase and management of prepaid virtual Visa cards usable at US-based online merchants. - Provides a CLI tool for signup, card creation, viewing card details, checking balances, and processing refunds. - All card data is securely stored server-side, encrypted with AES-256-GCM. - Includes detailed user flows for card purchase, checkout, balance tracking, and refunds. - US merchants only; cards work globally but must be used with US companies. - Built by Alchemy; package has passed internal and independent security audits.
元数据
Slug agentcardai
版本 1.0.8
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 8
常见问题

agentcard.ai 是什么?

Purchase and manage prepaid virtual Visa cards usable at US online merchants for secure AI agent payments via the agentcard CLI. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 110 次。

如何安装 agentcard.ai?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install agentcardai」即可一键安装,无需额外配置。

agentcard.ai 是免费的吗?

是的,agentcard.ai 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

agentcard.ai 支持哪些平台?

agentcard.ai 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 agentcard.ai?

由 florrdv(@florrdv)开发并维护,当前版本 v1.0.8。

💬 留言讨论