← 返回 Skills 市场
lukasvanuden

AgentRef

作者 Lukas van Uden · GitHub ↗ · v1.0.1 · MIT-0
cross-platform ✓ 安全检测通过
107
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install agentref
功能描述
Use AgentRef's REST API from OpenClaw or any HTTP runtime for merchant onboarding, programs, affiliates, conversions, flags, and payouts; authenticate with A...
使用说明 (SKILL.md)

AgentRef REST

Use this skill when the user wants to inspect or operate AgentRef over HTTP instead of MCP, especially from OpenClaw or other REST-first runtimes.

Connection contract

  • Base URL: https://www.agentref.co/api/v1
  • Every request needs Authorization: Bearer \x3CAGENTREF_API_KEY>
  • Never print, log, echo, or persist the raw API key in chat output, files, traces, or examples
  • Success responses use { data, meta }
  • Error responses use { error, meta }
  • When a request fails, surface the failing endpoint plus meta.requestId

First request

  1. Call GET /api/v1/me
  2. Read data.key.ownerType, data.key.keyType, data.key.scopes, and the returned owner/program context
  3. If ownerType is affiliate, do not use the merchant-admin endpoints below; say so clearly and only pivot to affiliate /api/v1/me/* reads if the user explicitly wants affiliate self-service

Credential guidance

  • Prefer a dedicated AgentRef key for this skill instead of reusing a broad personal key
  • Default to the smallest scope set that fits the task
  • For inspection-only use, prefer read-only merchant scopes
  • Only use write-capable scopes such as program changes, affiliate moderation, flag resolution, or payouts when the user explicitly needs those actions
  • If the user only needs affiliate self-service, use an affiliate key instead of a merchant key

Operating rules

  • Safe reads first: inspect current state with GET before proposing or sending writes
  • Never invent endpoints, fields, or state transitions; use only the curated references in this bundle
  • Ask for clear user confirmation before any action that creates, updates, archives, approves, blocks, resolves, or pays out
  • When a documented write supports it, send an Idempotency-Key header
  • Follow pagination from meta.nextCursor or page and pageSize; do not guess hidden pages
  • Reuse IDs already returned by prior reads instead of searching again

Reference map

Load only the section needed for the current task. Keep the active prompt lean.

安全使用建议
This skill appears coherent and limited to calling AgentRef's REST API. Before installing: (1) provide a dedicated AgentRef API key with the minimum scopes needed (read-only if you only want inspection); (2) be cautious granting write-capable scopes (payouts, blocking affiliates, onboarding completion) because those have real business impact—the skill asks for confirmation before writes, but you should verify intent; (3) there is no installable code, so the main risk is the API key itself—treat it like other service credentials. If you need the agent to act autonomously with high-impact write permissions, consider restricting the key or requiring manual approval for those actions.
功能分析
Type: OpenClaw Skill Name: agentref Version: 1.0.1 The agentref skill bundle provides a legitimate interface for managing affiliate programs via the AgentRef REST API (https://www.agentref.co/api/v1). It includes comprehensive documentation of endpoints and workflows in references/endpoints.md and references/workflows.md, with a strong emphasis on security best practices. Specifically, SKILL.md instructs the agent to protect the AGENTREF_API_KEY and require explicit user confirmation before performing any state-changing or destructive actions, such as payouts or blocking affiliates. No indicators of malicious intent, obfuscation, or unauthorized data exfiltration were identified.
能力标签
requires-oauth-token
能力评估
Purpose & Capability
Name/description and required items align: the skill targets AgentRef REST usage and only requests AGENTREF_API_KEY. No unrelated credentials, binaries, or config paths are requested.
Instruction Scope
SKILL.md limits actions to AgentRef endpoints in the bundled references, mandates GET /api/v1/me first, prefers safe reads, and requires explicit confirmation for writes. It does not instruct reading local files, other env vars, or contacting unexpected external endpoints.
Install Mechanism
Instruction-only skill with no install spec or code to write to disk; lowest install risk.
Credentials
Only AGENTREF_API_KEY is required and declared as the primary credential. The SKILL.md explicitly recommends using a dedicated, least-privilege key which is appropriate for the features exposed.
Persistence & Privilege
always:false and no install actions or modifications to other skills; the skill does not request permanent elevated presence or cross-skill config changes. Note: autonomous invocation (model calls the skill) is allowed by default but is expected for skills and not flagged here.
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 agentref
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /agentref 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
v1.0.1
AgentRef skill 1.0.1 - Added OpenClaw integration metadata, including homepage URL and required environment variable (AGENTREF_API_KEY). - Introduced guidance to prefer dedicated, minimally-scoped API keys for safer operation. - Clarified recommendations around credential usage, including details for inspection-only and affiliate-only flows. - No changes to functionality or endpoints; documentation improvements only.
v1.0.0
Initial public release of the AgentRef OpenClaw skill. Includes REST API authentication guidance, core AgentRef workflows, and curated endpoint references.
元数据
Slug agentref
版本 1.0.1
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 2
常见问题

AgentRef 是什么?

Use AgentRef's REST API from OpenClaw or any HTTP runtime for merchant onboarding, programs, affiliates, conversions, flags, and payouts; authenticate with A... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 107 次。

如何安装 AgentRef?

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

AgentRef 是免费的吗?

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

AgentRef 支持哪些平台?

AgentRef 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 AgentRef?

由 Lukas van Uden(@lukasvanuden)开发并维护,当前版本 v1.0.1。

💬 留言讨论