跳到正文
cccskills
免费在 GitHub 公开

build-ai-matrx-agents

Splits one big Claude task or automation into a set of focused AI Matrx agents that work on the person's own AI Matrx tables, then builds and test-runs each one. Use when the person says "split my task into agents", "turn my automation into AI Matrx agents", "make agents for my work", or after their data has moved into AI Matrx.

中文简介正在准备中。当前显示原文说明。

实测评分 未实测

查看安装方法

包含的文件(1)

  • SKILL.md2.6 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Turn one big task into AI Matrx agents

The person has one large Claude task (a long prompt, a CLAUDE.md, a project instruction or a scheduled job) that does many jobs at once. You split it into about a dozen focused AI Matrx agents, each doing one job well on their AI Matrx data, owned by them in their organization.

1. Read what exists

  1. Read their big task in full (ask where it lives if you do not have it).
  2. tables action: "list_tables", then action: "columns" on each table the task touches, so every agent is described in terms of their real tables and columns.
  3. agents action: "list" — reuse an agent they already have rather than making a twin.

2. Propose the split

One agent per job a person would name on its own ("write next week's captions for a client", "flag invoices more than 15 days overdue", "summarize a campaign's results for the client"). For each, write down — in plain words, never as a prompt:

  • Name — Title Case, what it does ("Weekly Caption Writer").
  • Goal — what it is for and what a good result looks like.
  • Inputs — the variables each run is given (client_name, week_of, …), each with a description.
  • Output — text, or a JSON shape when its result feeds a table or another agent.
  • Tables — which tables it reads and which it changes.

Show the person the list as one short line per agent and ask once: "Shall I build these?"

3. Build and prove each one

For each agent:

agents action:"create" name:"Weekly Caption Writer" goals:"…" summary:"…"
       variables:[{"name":"client_name","description":"…"}] response_format:"text"

AI Matrx's Agent Builder writes the agent from the goal (a minute or two each); agents get the tools to read and change the person's tables by default. Then run it once on real data:

agents action:"run" agent_id:"…" values:{"client_name":"Fernwood Bakery"}

Check the output against the goal. If it misses, say what was missing and rebuild by calling create again with a clearer goal (never write its prompt yourself).

4. Hand over

Give the person: each agent's name, one line on what it does, and the result of its test run. Tell them where they live: AI Matrx → Agents. Note anything the old big task did that no agent covers yet.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Admin debug system wiring for the floating AdminIndicator's live debug data and Copy Full Context. Use when adding debug visibility to a route or feature, wiring useDebugContext or the debug panel, capturing console errors, or enabling the copy-context workflow for a page.

中文简介正在准备中。当前显示原文说明。

armanisadeghi/ai-matrx42026年10月23日 更新

Compact two-icon Copy / Copy-for-AI controls (components/agent-copy). Use when adding copy buttons to a row, card, list, or record; merging duplicate Copy/JSON/AI controls; continuing the copy rollout; or writing a Copy-for-AI payload. NOT for markdown content actions (use rich-document-actions).

中文简介正在准备中。当前显示原文说明。

armanisadeghi/ai-matrx42026年10月23日 更新

Disclosing a surface's existing fixed AI jobs in the shell's top Agents menu. Use when a page, panel, overlay, or window already runs a mandate behind a button, assist, automatic action, or mode; when the agent-disclosure guard names a file; or during a surface check.

中文简介正在准备中。当前显示原文说明。

armanisadeghi/ai-matrx42026年10月23日 更新

The method for getting a cheaper model to match an agent's big-model answers by fixing instructions, inputs and setup. Use when asked to move an agent to a smaller or cheaper model, prove one model beats another, or compare models on an agent's test cases.

中文简介正在准备中。当前显示原文说明。

armanisadeghi/ai-matrx42026年10月23日 更新

The watch-fix-rerun method for making an existing platform agent or automated process efficient and correct: baseline its ledger, run one unit yourself, fix the class behind every wasted call, rerun, record. Use when asked to improve, optimize, watch, or 'make efficient' an agent, a sandbox session, a sync job, or any recurring automated process.

中文简介正在准备中。当前显示原文说明。

armanisadeghi/ai-matrx42026年10月23日 更新

Redux state for agent execution and firing agent shortcuts. Use when editing features/agents/redux/, building agent UI, creating a conversation, touching assembleRequest or NDJSON stream state, wiring agent-state selectors, adding a per-conversation capability, or triggering a shortcut from a button, menu, or code (useShortcutTrigger, launchAgentExecution).

中文简介正在准备中。当前显示原文说明。

armanisadeghi/ai-matrx42026年10月23日 更新

查看 armanisadeghi 的全部技能

报告此技能的问题