What makes this different from regular AI documentation tools?

This skill operates on a 'strict truth' basis. It requires concrete evidence from your codebase for every claim it makes and explicitly marks gaps as 'To be confirmed' instead of hallucinating details.

Can I use this for Architecture Decision Records (ADR)?

Absolutely. It includes built-in naming conventions and templates for ADRs, PRDs, and integration guides that follow industry best practices.

Does it support incremental updates to existing docs?

Yes, it prioritizes a patch-diff mode that allows it to update specific sections of your documentation without rewriting or overwriting your entire file.

How does it handle sensitive information like API keys?

The skill is strictly configured to never output plaintext secrets. It uses placeholders and provides the necessary commands or paths for users to retrieve those values securely.

DDD Document Steward

Name: DDD Document Steward
Author: tianzecn

bytianzecn

•

コンテンツ管理

Automates the maintenance of Single Source of Truth documentation by synchronizing project docs with verifiable codebase evidence.

DDD Document Steward is a specialized Claude Code skill designed to implement Documentation-Driven Development (DDD) by ensuring your project's documentation remains the Single Source of Truth (SSOT). Unlike standard AI writers, this skill operates under a 'strict' truthfulness mode, refusing to speculate on architectural details. It follows a rigorous four-phase workflow—Inventory, Planning, Generation, and Consistency Checking—to produce high-quality guides, architecture decision records (ADRs), and feature specifications. By scanning your actual codebase, configuration files, and routes, it creates documentation that is always backed by evidence, marking any unverified information as 'To be confirmed' with clear paths for manual validation.

主な機能

01Incremental updates using patch diffs to preserve existing manual documentation

021 GitHub stars

03Automated documentation inventory and gap analysis across project directories

04Automated 'To be confirmed' flagging for missing data with actionable verification steps

05Evidence-based documentation generation linked to specific file paths and configurations

06Strict adherence to standard structures including Purpose, Scope, Evidence, and Changelogs

ユースケース

01Establishing a verifiable Single Source of Truth (SSOT) for legacy repositories

02Syncing architectural records and integration guides during rapid feature development

03Generating incident post-mortems and technical specifications based on Git diffs

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tianzecn/myclaudecode ddd-doc-steward

For use in Claude.ai and ChatGPT

Download Skill