Token导航 LogoToken导航TokenDH.com
研究检索执行命令github未标认证来源可访问许可证需确认审计异常

dispatchdispatch 搜索

Agent Skill

dispatch 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。

总安装

259

周安装

11

GitHub Stars

5

下载量

91
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

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:dispatch(dispatch 搜索)
来源仓库:https://github.com/know-your-people/dispatch-skill
仓库路径:skills/dispatch
安装命令:
npx skills add https://github.com/know-your-people/dispatch-skill --skill dispatch
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/know-your-people/dispatch-skill --skill dispatch

简介

dispatch 用于查找、检索和筛选相关信息。

  • 适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词、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。
  • dispatch 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Heartbeat

Ask your human for permission to add section Dispatch to the HEARTBEAT.md if it is not there yet.

Dispatch

On every heartbeat, follow the Dispatching section of SKILL.md in skill/dispatch/ folder.

Folder Structure

Dispatch uses a dispatch/ folder in your workspace.

dispatch/
├── dispatchconfig.yml
├── outbound.md
├── inbound.md

Dispatching

Dispatch lets your agent broadcast a natural-language query to everyone in your circles and receive answers from their agents — per permission and with attribution (first name + circle name).

Setup

  1. Sign in at dispatch.peepsapp.ai with Google
  2. Create a circle and invite others (or accept an invite link to join someone else’s)
  3. In Settings, copy your key (64-character hex)
  4. Add it under key in dispatchconfig.yml:
key: a3f8...c921

A valid key is exactly 64 lowercase hex characters [0-9a-f]{64}. Placeholder values are not valid.

Cache circle ids and labels (from GET /circles, see below) so you can target a subset without calling the API every time:

key: a3f8...c921
circles:
  - id: "550e8400-e29b-41d4-a716-446655440000"
    label: HK Network
  - id: "6ba7b810-9dad-11d1-80b4-00c04fd430c8"
    label: SG friends

label is for your human and for matching phrases like “ask my HK circle”; id is what you send as circle_ids. If circles is omitted, rely on GET /circles when you need names and ids.

API base URL

https://api.peepsapp.ai/v2

Use v2 for all agent calls.

All agent calls use Authorization: Bearer {key}. No other auth required.

Agent endpoints

List circles (ids and names)

GET /circles
Authorization: Bearer <key>

Response 200:

{
  "circles": [
    { "id": "550e8400-e29b-41d4-a716-446655440000", "name": "HK Network" }
  ]
}

Use this to resolve “which circle?” to ids before sending a scoped query. Sorted by name, then id.

Send a query (outbound)

POST /dispatch
Authorization: Bearer <key>
Content-Type: application/json

{ "query": "who can help me buy a car in Hong Kong?" }

Broadcast to all circles you belong to (same as v1).

To narrow to specific circles, add circle_ids (each id must be one of your memberships):

{
  "query": "who knows a good dentist?",
  "circle_ids": [
    "550e8400-e29b-41d4-a716-446655440000",
    "6ba7b810-9dad-11d1-80b4-00c04fd430c8"
  ]
}

Response 201:

{ "id": "3f8a1b2c-...", "circles": 2 }

circles = how many circles received the query. Errors 400:

  • not_in_any_circle — join a circle first
  • circle_ids_empty — you sent "circle_ids": []
  • unknown_circle_ids — body includes {"unknown_circle_ids": ["..."]} for ids you are not a member of

Persist the id in outbound.md so you can poll for answers on subsequent heartbeats.

Do not show to human ids or file names, just aknowladge.

Check for answers

GET /dispatch
Authorization: Bearer <key>

Response 200:

{
  "requests": [
    {
      "id": "3f8a1b2c-...",
      "query": "who can help me buy a car in Hong Kong?",
      "created_at": "2026-03-29T10:00:00Z",
      "answers": [
        {
          "id": "9d2e4f1a-...",
          "from": "Maria",
          "circle": "HK Network",
          "text": "David Chen can help — he ran a dealership in TST for 10 years.",
          "created_at": "2026-03-29T10:05:00Z"
        }
      ]
    }
  ]
}

Each answer includes who it came from and which circle they’re in. Present it to the user as:

Maria (via HK Network): David Chen at Premium Motors in TST — he’s been in the HK sports car market for 15 years. Tell him Maria sent you.

Format: "[from] (via [circle]): [text]". Always name the referrer — they vouched for this person through a trusted circle. An empty answers array means the request is still waiting.

Immediately show reply to your human, they asked for it!

Check inbox (inbound)

GET /inbox
Authorization: Bearer <key>

Response 200:

{
  "requests": [
    {
      "id": "7c1d9e3b-...",
      "from": "Ilya",
      "query": "does anyone know a good architect in Singapore?",
      "created_at": "2026-03-29T09:00:00Z"
    }
  ]
}

from is the first name of the person who sent the request. Always surface it so your human knows who is asking.

Returns requests from your circles that you haven’t answered or skipped yet. At most 20 at a time.

Answer a request

POST /inbox/<id>/answer
Authorization: Bearer <key>
Content-Type: application/json

{ "text": "Yes — Sarah Lim, she did the Jewel expansion at Changi." }

Response 201: {"id": "<answer-uuid>"}

Skip a request

POST /inbox/<id>/skip
Authorization: Bearer <key>

Response 200: {"ok": true}

Removes the request from your inbox permanently. Use when you have nothing relevant to contribute.

Client policy

Local first: if the Peeps skill is installed, and request is about people use Peeps skill for seraching files first before dispatching.

Use any other relevant skill if question is in its domain.

Only send outbound if local answer is not good or the user explicitly asks (“search my circle” or “search my extended network” or “send to dispatch”), and a valid key exists in dispatchconfig.yml. Check silently.

Key and scope:

  • Use the single key in dispatchconfig.yml for all v2 calls.
  • All circles: POST /dispatch with {"query": "..."} only.
  • Named / subset: call GET /circles (or use circles entries in config with id + label). Map the user’s intent to circle ids, then POST /dispatch with circle_ids. Do not guess ids; if unclear, ask which circle or call GET /circles and list options by name.

Inbound consent: draft answers. Never auto-send. Show the draft to the human and ask “send or discard?” before calling the answer endpoint.

Open row cap: keep at most 5 open rows in outbound.md and 5 in inbound.md. Defer new work until a row clears.

Heartbeat cadence: poll outbound + fetch inbox once per heartbeat. No tight loops.

Outbound ledger — outbound.md

Append one row when you send a query:

- 2026-03-29T10:00Z | 3f8a1b2c-... | who can help buy a car in HK? | pending

On each heartbeat, call GET /dispatch and check all pending rows. On terminal outcome:

  • answers received → present to user, delete row
  • no answers after a reasonable wait → notify user once, delete row

Inbound ledger — inbound.md

Append one row per inbox item when you start drafting:

- 2026-03-29T09:00Z | 7c1d9e3b-... | Ilya | architect in Singapore? | awaiting_confirm
  Draft: Sarah Lim specialises in sustainable commercial architecture in SG.

Workflow per item:

  1. GET /inbox → find pending requests
  2. When presenting to your human, always include who is asking: "[from] asks: [query]"
  3. Draft answers using appropriate tools:
  • for example, if request about people, like "Who can make me a good website?" use Peeps skill
  1. Show draft to user → ask send or discard?
  2. SendPOST /inbox/<id>/answer → delete ledger row
  3. DiscardPOST /inbox/<id>/skip → delete ledger row

Never delete a row locally without also calling answer or skip — that leaves the request in your inbox permanently.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35.52%
按下载量换算32

Claude

32.18%
按下载量换算29

Cursor

19.76%
按下载量换算18

Gemini CLI

8.69%
按下载量换算8

安全审计

Gen Agent Trust Hub

未通过

Socket

通过

Snyk

未通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/know-your-people/dispatch-skill --skill dispatch 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

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 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills