这是一项为编码代理准备的技能，可以将Filemaker数据库设计报告转化为迁移计划。

FileMaker 将 DDR 导出为每个数据库文件中的一个大型 XML 文件。一个 300 MB 的 DDR 包含数千万个令牌，远远超出代理上下文的容量。因此，技能首先将 DDR 拆分为每个对象的一个文本文件：每个脚本、表格、布局、关系、值列表、自定义函数、菜单、账户和权限集。代理读取这些文件，分类，并根据它们的描述写入计划。计划中的每个申报都指向一个对象文件。

你需要什么

• 加载技能的编码代理（Claude Code、Codex 或类似代码）
• Python 3。没有包。
• 完整的DDR XML导出：每个FileMaker文件一个XML文件，如果有Summary.xml，还要加

DDR可以以明文形式保存密码和API密钥。爆炸文件会对它们进行涂黑;原始XML则不会。

安装

把仓库克隆到你代理的技能文件夹里。然后要求$migrate-filemaker，或者让代理加载SKILL.md。

它是如何运作的

1. 提取。explode_ddr.py 为每个对象写一个文本文件，包含所有对象的 index.tsv，代码的 LEGEND.md 和一个“未解决”。

migrate-filemaker

English | 日本語

A skill for coding agents that turns a FileMaker Database Design Report into a migration plan.

FileMaker exports a DDR as one large XML file per database file. A 300 MB DDR is tens of millions of tokens, far more than fits in an agent’s context. So the skill first breaks the DDR into one text file per object: every script, table, layout, relationship, value list, custom function, menu, account and privilege set. The agent reads every one of those files, classifies it, and writes the plan from what they say. Every claim in the plan points at an object file.

Migration examples

Migration projects that use this skill.

• migrate-karbonfm: migration of KarbonFM, an open-source FileMaker CRM, to a desktop app for macOS, Windows and Linux.

What you need

• A coding agent that loads skills (Claude Code, Codex, or similar)
• Python 3. No packages.
• The full DDR XML export: one XML file per FileMaker file, plus Summary.xml if you have it

The DDR is client data, and the exploded files carry every value exactly as the DDR has it. Keep both out of version control.

Install

Clone the repo into your agent’s skills folder. Then ask for $migrate-filemaker, or tell the agent to load SKILL.md.

How it works

1. Extract. explode_ddr.py writes one text file per object, an index.tsv of all of them, a LEGEND.md for the codes, and an unresolved.tsv of references to deleted or missing targets. It stops if any calculation or attribute check fails.
2. Inventory. Count what is there and how the files connect.
3. Read every object. Nine explorers (scripts, tables, layouts, graph, security, value lists, menus, custom functions, file settings) each read their kind and write one CSV row per object. catalog_contract.py verify proves nothing was skipped.
4. Ask. List what the DDR cannot answer: goals, users, identity, integrations, which behaviors to keep.
5. Plan. Write the schema, API, UI, business logic, authorization, data migration and cutover documents from the templates.

The skill records what the FileMaker solution does. It does not judge whether that logic is right.

What an object file looks like

A script reads the way Script Workspace shows it, with the ids a search needs. // marks a disabled step.

# script 12: Go Home
# file: Orders.fmp12