Token导航 LogoToken导航

Agent Skill · 研究检索

airtop-agents(空中顶 Agent)

airtop-agents 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

airtop-agents 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 58,684。

总安装 408周安装 17GitHub Stars 1下载 136核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:airtop-agents(空中顶 Agent)
来源仓库:https://github.com/airtop-ai/airtop-skill
仓库路径:skills/airtop-agents
安装命令:
npx skills add https://github.com/airtop-ai/airtop-skill --skill airtop-agents
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/airtop-ai/airtop-skill --skill airtop-agents

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

airtop-agents 用于通过 REST API 列出、运行、监控和取消 Airtop 平台的 AI Agent 实例。适合需要远程管理 Airtop 服务中代理任务的用户,支持状态跟踪和生命周期控制。必须配置 AIRTOP_API_KEY 环境变量或通过 .env 文件提供认证凭据方可使用。

操作涉及外部 API 调用,需注意密钥安全,避免在公开对话中泄露敏感信息。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/airtop-agents

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/airtop-ai/airtop-skill --skill airtop-agents;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

airtop-agents 怎么安装?

在终端执行 npx skills add https://github.com/airtop-ai/airtop-skill --skill airtop-agents。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

airtop-agents 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

airtop-agents 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

airtop-agents 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

airtop-agents 用的人多吗?

公开数据是累计安装 408 次,近一周安装 17 次,下载 136 次,GitHub 1 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

airtop-agents 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
airtop-agents

空中顶 Agent · 本页

4081接触敏感数据未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Airtop Agents Skill

You can list, run, monitor, and cancel Airtop agents using their REST API.

Authentication

The Airtop API key is required for all operations. Resolve it in this order:

  1. $AIRTOP_API_KEY environment variable
  2. A .env file in this skill's directory containing AIRTOP_API_KEY=...
  3. If neither is found, offer the user two options before proceeding:
Option A — Set it up yourself (recommended if you prefer not to share your key in chat): Run these commands in your terminal: `` cp "$(dirname "$SKILL_PATH")/.env.example" "$(dirname "$SKILL_PATH")/.env" ` Then open the .env file and replace your-api-key-here with your key from [https://portal.airtop.ai/api-keys](https://portal.airtop.ai/api-keys). Once done, say "done" and I'll pick it up automatically. **Option B — Paste it here and I'll save it for you:** Paste your API key (from [https://portal.airtop.ai/api-keys](https://portal.airtop.ai/api-keys)) and I'll write it to the .env` file so it's available for future use.

Print both options exactly as above (with the actual resolved path instead of the $(dirname...) expression) and wait for the user to choose. Do not assume a preference.

Handling a pasted key (Option B)

Important — always load the key from the .env file, never use a pasted value directly.

When a user provides their API key interactively (e.g. pasting it into chat), text copied from web UIs or chat messages can contain invisible Unicode characters (zero-width spaces, byte-order marks, etc.) that silently break authentication. To avoid this:

  1. Write the key to .env first — this round-trips it through file I/O which strips invisible characters: echo "AIRTOP_API_KEY=<pasted-value>" > "$(dirname "$SKILL_PATH")/.env"
  2. Then read it back from the file to get a clean value: API_KEY=$(grep AIRTOP_API_KEY "$(dirname "$SKILL_PATH")/.env" | cut -d= -f2-)

Even when $AIRTOP_API_KEY is already set in the environment, prefer loading from .env if the file exists — the environment variable may have been set in the same shell session from a pasted value and could carry the same invisible characters.

Never assign a user-pasted key directly to a shell variable and use it in API calls (e.g. API_KEY="<pasted-value>" followed by curl -H "Authorization: Bearer $API_KEY"). Always go through the .env file write-then-read cycle to sanitize the value.

Loading and validating the key

Once the .env file exists (whether set up by the user or written by you), load and validate:

API_KEY=$(grep AIRTOP_API_KEY "$(dirname "$SKILL_PATH")/.env" | cut -d= -f2-)

Validate the key immediately after loading it:

curl -sf -H "Authorization: Bearer ${API_KEY}" "https://api.airtop.ai/api/v2/agents?limit=1" > /dev/null

If this returns a non-zero exit code, tell the user their API key appears invalid and link them to https://portal.airtop.ai/api-keys.

Base URL

All authenticated API endpoints use: https://api.airtop.ai/api

Webhook endpoints (run agent, poll result) use the public path: https://api.airtop.ai/api/hooks/

Argument Parsing

Parse $ARGUMENTS as follows:

list [--name <filter>]              List agents
run <agent-name-or-id> [--vars {}]  Run an agent with optional config variables
status <agentId> <invocationId>     Check invocation status
cancel <agentId> <invocationId>     Cancel a running invocation
history <agentId>                   View recent invocations

If $ARGUMENTS is empty or unrecognized, show the usage summary above and ask the user what they'd like to do.

Commands

1. List Agents

curl -s -H "Authorization: Bearer $API_KEY" \
  "https://api.airtop.ai/api/v2/agents?limit=25&sortBy=lastRun&sortOrder=desc&createdByMe=true"

Always include createdByMe=true to show only agents owned by the current user (not the entire organization).

Optional query parameters to append:

  • &name=<filter> — case-insensitive partial match (use when --name is provided)
  • &enabled=true — filter to enabled agents only
  • &published=true — filter to published agents only

Display: Format the response as a markdown table with columns: Name, ID, Enabled, Last Run. Show the pagination.total count in a header line.

Example output:

Found 3 agents:
| Name              | ID                                   | Enabled | Last Run    |
|-------------------|--------------------------------------|---------|-------------|
| Price Tracker     | 550e8400-e29b-41d4-a716-446655440000 | Yes     | 2 hours ago |
| Lead Enricher     | 6ba7b810-9dad-11d1-80b4-00c04fd430c8 | Yes     | Yesterday   |
| Competitor Monitor| 6ba7b811-9dad-11d1-80b4-00c04fd430c8 | No      | Never       |

2. Run Agent

This is a multi-step process:

Step 1 — Resolve agent by name or ID.

If the argument looks like a UUID, use it directly as the agent ID. Otherwise, search by name:

curl -s -H "Authorization: Bearer $API_KEY" \
  "https://api.airtop.ai/api/v2/agents?name=$(printf '%s' '<agent-name>' | jq -sRr @uri)&createdByMe=true"
  • If exactly one agent matches, use it (but still validate it in Step 2).
  • If multiple agents match, prefer enabled and published agents over disabled or draft-only ones. If there is still ambiguity, display them in a table (with Enabled and Published status columns) and ask the user to pick one.
  • If no agents match, tell the user no agent was found and suggest running list.

Step 2 — Validate the agent is runnable.

Fetch the full agent details (already available from step 1 if resolved by name, or fetch now):

curl -s -H "Authorization: Bearer $API_KEY" \
  "https://api.airtop.ai/api/v2/agents/<agentId>"

Check the following before proceeding:

  • Disabled agent (enabled is false): Tell the user the agent is disabled and cannot be run. Suggest they enable it in the Airtop portal.
  • Draft-only agent (hasDraft is true AND publishedVersion is absent): Tell the user the agent has only a draft version and has never been published. It must be published in the Airtop portal before it can be invoked via webhook.
  • Published agent with draft (hasDraft is true AND publishedVersion is present): The agent is runnable. Note to the user that the agent has unpublished draft changes, and the published version will be used. Use the publishedVersion value for the version parameter.
  • Published agent (publishedVersion is present, hasDraft is false): The agent is runnable. Use the publishedVersion value for the version parameter.

Step 3 — Check required configVars.

The agent details response includes a versionData.configVarsSchema field. Inspect it for required properties. If the user has not provided values for required configVars via --vars, prompt them for the missing values before invoking.

Display the required and optional parameters with their descriptions and defaults so the user knows what to provide.

Step 4 — Get the agent's webhook.

curl -s -H "Authorization: Bearer $API_KEY" \
  "https://api.airtop.ai/api/v2/agents/<agentId>/webhooks?limit=10"

Use the first webhook from the webhooks array. If no webhooks exist, tell the user the agent needs a webhook configured in the Airtop portal (https://portal.airtop.ai).

Step 5 — Invoke the webhook.

curl -s -X POST "https://api.airtop.ai/api/hooks/agents/<agentId>/webhooks/<webhookId>" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "configVars": { <user-provided key-value pairs from --vars, or empty object {}> },
    "version": <publishedVersion from agent details>
  }'

Always use the publishedVersion value from the agent details as the version parameter — never hardcode it.

The response contains {"invocationId": "<uuid>"}. Save this for polling.

Step 6 — Poll for the result.

curl -s "https://api.airtop.ai/api/hooks/agents/<agentId>/invocations/<invocationId>/result" \
  -H "Authorization: Bearer $API_KEY"

Polling rules:

  • Poll every 5 seconds
  • Print a status update to the user after each poll (e.g., "Status: Running...")
  • Terminal statuses: Completed, Failed, Cancelled
  • Timeout after 5 minutes (60 polls). If not complete, tell the user the invocation is still running and provide the invocation ID so they can check later with the status command.

Step 7 — Display the result.

  • On Completed: Show the output field. If it's JSON, format it nicely. If it's a string, display it directly.
  • On Failed: Show the error field and the full status.
  • On Cancelled: Inform the user the invocation was cancelled.

3. Check Status

Poll a specific invocation's result:

curl -s "https://api.airtop.ai/api/hooks/agents/<agentId>/invocations/<invocationId>/result" \
  -H "Authorization: Bearer $API_KEY"

Display the status field. If terminal, also display output or error.

Note: This requires both the agent ID and invocation ID. If the user only provides one, ask for the other.

4. Cancel Invocation

curl -s -X DELETE -H "Authorization: Bearer $API_KEY" \
  "https://api.airtop.ai/api/v2/agents/<agentId>/invocations/<invocationId>?reason=user_requested"

Confirm to the user that the cancellation was requested.

5. View History

curl -s -H "Authorization: Bearer $API_KEY" \
  "https://api.airtop.ai/api/v2/agents/<agentId>/invocations?limit=10"

Display results as a table with columns: Invocation ID, Status, Trigger, Started At.

Error Handling

  • 401 Unauthorized: Tell the user their API key is invalid or expired. Direct them to https://portal.airtop.ai/api-keys.
  • 404 Not Found: The agent or invocation doesn't exist. Suggest checking the ID or running list.
  • 429 Rate Limited: Tell the user they've hit the rate limit and should wait before retrying.
  • No webhook configured: Explain that the agent needs a webhook set up in the Airtop portal before it can be invoked from the CLI.
  • Multiple name matches: List all matches and ask the user to pick one or use the agent ID directly.
  • Empty API key: Guide the user to set AIRTOP_API_KEY or provide it interactively.

Important Notes

  • Always use curl -s (silent mode) to avoid progress bars in output.
  • Parse all JSON responses with jq or inline JSON parsing in bash.
  • The webhook invoke and poll endpoints are under /api/hooks/ (public path), NOT /api/v2/.
  • The list, webhooks, history, and cancel endpoints are under /api/v2/ (authenticated path).
  • When displaying times, convert ISO timestamps to human-readable relative times (e.g., "2 hours ago").

平台分布

Codex

39.12%
按下载量换算53

Claude

28.17%
按下载量换算38

Cursor

19.36%
按下载量换算26

Gemini CLI

9.75%
按下载量换算13

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills