- Home
- Skills
- Data & Databases
- Database Schema Handbook: Data Dictionary and ERD
More screenshots
Works with the AI Agents you already use
Database Schema Handbook: Data Dictionary and ERD
The agent drafts a schema handbook, for you to review: a data dictionary, a diagram, the relationships in words and a list of findings.
$7
Database Schema Handbook: Data Dictionary and ERD
Example session with this skill installed
Document this database. schema.sql is a pg_dump from production of our clinic booking system; a new developer starts Monday and needs to find their way around.
- Read your context and instructions
- Compiled the database schema handbook
appointment_status_log
- Purpose (inferred): One row per change of an appointment's status, with the status before and after. The table has no primary key.
- Source:
input/schema.sqlline 67
| Column | Type | Null | Key | Meaning | Basis |
|---|---|---|---|---|---|
appointment_id | bigint | no | FK -> appointments.id | The appointment whose status changed; its log rows are deleted with it. | constraint |
changed_by | integer | yes | [NEEDS: Whose id is this, a staff.id or something else? No foreign key is declared.] | unknown |
Summary
Observations: 35, of 13 kinds, across 15 tables. Open questions: 33. Statements not read: 4.
Foreign key type differs from the referenced column (1)
invoices.patient_idisinteger;patients.idisbigint-input/schema.sqlline 542
Connects securely to your tools. The creator never sees your data.
About this skill
English
For developers, agencies and teams who inherit a database, hand one over, or have to explain one to a new colleague or a client.
Give your agent the database's own schema files. The agent drafts a schema handbook, for you to review: a data dictionary, a diagram, the relationships in words and a list of findings.
It reads a schema.sql, a pg_dump --schema-only file, a MySQL dump, an SQLite schema or a folder of SQL migrations applied in order. Nothing connects to a database.
Meanings the agent infers are labelled as inferred and stay unverified until your team confirms them. Views, functions and triggers are listed, not described.
What the agent writes each run
data-dictionary.mdanddata-dictionary.csv: every table and column with type, nullability, default and keys, a plain-language meaning, and the basis of that meaning: schema comment, constraint, inferred from the name, told by you, or unknown.erd.mmd: a Mermaid diagram to paste into a wiki or a README.relationships.md: each foreign key in words with its source line, many-to-many links, and join examples for the most connected tables.findings.md: factual observations, each naming the table or column and citing file and line. Examples: tables with no primary key, foreign keys with no supporting index, type mismatches across a foreign key, money-like columns stored as floating point, date-time columns with no time zone, duplicate indexes.handbook.html: one self-contained page with the diagram drawn inline, every table, the relationships, the findings and the open questions. It prints as is.
Read by a script, explained by the agent
A bundled script parses the DDL for PostgreSQL, MySQL and SQLite, applies ALTER statements and migrations in order, and records the file and line of every table, column, key and index. The agent then writes what each table and column means. A check command refuses a dictionary with a missing table or column, a stale type or a blank meaning.
Your schema only
The skill instructs the agent to label every inferred meaning, and to write what it cannot know as a [NEEDS: ...] question for your team instead of filling it in.
Honest about its limits
It documents structure as declared in the files. Row counts, data quality, index usage and what the application does at run time are not in those files, so they are not in the handbook. Meanings inferred from names are drafts to confirm. Views, functions, procedures, triggers and exclusion constraints are listed with their lines and not described. The parser is tolerant, not complete: SQL Server and Oracle files were not tried. Name-based findings are heuristics and each may be a deliberate choice. Reading ORM models (Prisma, Django, Rails) is done by the agent with a template and was not run as a test case. The HTML page and its diagram were checked as markup, not viewed in a browser; on a large schema the diagram is tall and lines can cross.
Arabic
On request the handbook is written in Arabic with a right-to-left page; table and column names stay left to right.
In the download: SKILL.md, the standard-library Python scripts, references (reading schemas, writing meanings, the findings guide, Arabic notes), templates, and a worked example you can read before you buy: a 15-table PostgreSQL clinic booking database with 116 columns and 19 foreign keys, giving 35 observations and 33 open questions.
Tried on two cases (a PostgreSQL dump in English; a MySQL shop built from 8 migrations, in Arabic). The scripts have 52 unit tests.
بالعربي
للمطوّرين والوكالات والفرق التي تستلم قاعدة بيانات أو تسلّمها أو تحتاج أن تشرحها لزميل جديد أو لعميل.
أعطِ وكيلك ملفات مخطط قاعدة البيانات (schema.sql أو ملف تفريغ أو مجلد ملفات الترحيل). يكتب لك مسودة دليل لتراجعها: قاموس بيانات، مخطط علاقات، شرح العلاقات، وقائمة ملاحظات. لا يتصل بقاعدة البيانات إطلاقًا.
ماذا يكتب الوكيل
- ملف
data-dictionary: كل جدول وكل عمود مع نوعه ومفاتيحه ومعناه، ومصدر هذا المعنى. - ملف
erd.mmd: مخطط علاقات بصيغة Mermaid. - ملف
relationships.md: كل علاقة مشروحة بالكلمات مع أمثلة ربط. - ملف
findings.md: ملاحظات واقعية، مع كل ملاحظة اسم الملف ورقم السطر. - ملف
handbook.html: صفحة واحدة جاهزة للطباعة.
حدودها
توثّق البنية كما هي مكتوبة في الملفات فقط، فلا ترى البيانات نفسها ولا ما يفعله التطبيق. المعاني المستنتجة من الأسماء مسودات تحتاج تأكيدك، وما لا يُعرف يُكتب سؤالًا لفريقك. تدعم PostgreSQL و MySQL و SQLite. يمكن طلب الدليل بالعربية؛ جُرّب ذلك على حالة اختبار واحدة.
It documents structure as declared in the files. Row counts, data quality, index usage and run-time behaviour are not in those files. Meanings inferred from names are drafts to confirm. Views, functions, procedures, triggers and exclusion constraints are listed with their lines and not described. The parser is tolerant, not complete: SQL Server and Oracle files were not tried. Name-based findings are heuristics and each may be a deliberate choice. Reading ORM models (Prisma, Django, Rails) is done by the agent with a template and was not run as a test case. The HTML page and its diagram were checked as markup, not viewed in a browser; on a large schema the diagram is tall and lines can cross. Tried on two cases. المعاني المستنتجة من الأسماء مسودات تحتاج تأكيدك.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 2 days ago
Requires
Permissions
Reads the schema or migration files you name. Writes data-dictionary.md, data-dictionary.csv, erd.mmd, relationships.md, findings.md, handbook.html, schema.json and README.md into a folder in your project (schema-handbook by default). Runs its bundled Python scripts locally; they use the standard library only, make no network calls and never connect to a database. يقرأ ملفات المخطط ويكتب ملفات الدليل داخل مجلد في مشروعك فقط.