360 Secure SkillHub
360 安全技能中心

Feishu Identity Routing

为多代理、多账号场景构建飞书跨应用身份主数据:按 union_id、user_id 优先合并同一用户,把 app 级 open_id 归属到统一主体,并在出站动作前解析目标 accountId 对应的 open_id 以构造路由目标;附带单条合并、批量合并与待审批合并审阅脚本。

wxfkuq2023
wxfkuq2023
热度 33.89
最近更新 2026-05-11 17:38:19
分类:自动化依赖:代码执行 / 读写文件应用场景:工作流 / 自动化脚本 / 数据处理

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add feishu-identity-routing
Metadata
name
feishu-identity-routing
description
Build and use a Feishu/Lark cross-app identity master for multi-agent, multi-account routing. Use when mapping the same user across different Feishu app open_id values, resolving app-specific open_id from union_id/user_id, routing outbound actions with the correct accountId, or setting up reusable identity merge/review workflows.

Feishu Identity Routing

Use this skill when a workspace needs a reusable way to:

  • merge the same human across multiple Feishu apps/accounts
  • store app-local open_id under one canonical user subject
  • resolve the right open_id for a target accountId / app_context
  • route outbound actions correctly in multi-agent setups

What to create

Create these workspace files if missing:

  • identity/feishu-user-master.json — canonical data store
  • identity/feishu-user-master.md — human-readable summary

Core rules

  1. One person = one subject
  2. Merge by union_id first, then user_id
  3. Treat open_id as app-local only
  4. Before any outbound action:
    • resolve the user subject globally
    • resolve the target app/account-specific open_id
    • build the provider target from that local identity

For Feishu DM routing, use:

  • accountId=<target account>
  • target=user:<open_id>

Scripts

Use bundled scripts when you need deterministic updates:

  • scripts/merge_feishu_identity.js — merge one record
  • scripts/merge_feishu_identity_batch.js — merge many records
  • scripts/review_feishu_pending.js — approve/reject pending merges

When to read references

  • Read references/workflow.md when setting up or extending the process
  • Read references/schema-example.json when creating a new master file
  • Read references/pending-review-policy.md when conflicts appear
  • Read references/outbound-routing-patterns.md when implementing outbound delivery or cross-agent routing
  • Read references/examples.md when you want concrete merge/routing examples

Notes

  • Do not hardcode a specific agent like main or infoIntelOfficer into the design.
  • Model routing as: global subject resolution → app-context identity selection → provider-specific outbound target.
  • Keep the skill reusable across future agents and Feishu accounts.