chrome-ext-base
A Chrome extension starting point. Session-start auto-prepend, a commit gate, and a stop nag run on their own so AI doesn’t drift. Open the manual
ddakit
Same mistake twice and it stops on its own. When the work breaks mid-flow, one command brings you back. Code, prose, video, or legal documents — AI breaks in the same places.
The seven bases ddakit ships
Section titled “The seven bases ddakit ships”Three code bases, three writing and video studios, one legal harness. The surfaces differ, but the skeleton is the same: a safety net planted where AI breaks.
saas-base
A Vite + Supabase + Stripe SaaS starting point. FSD’s six-layer folder rules, a phase flow, and the AI entry points are all wired up in advance. Open the manual
e2e-base
An end-to-end testing starting point for external apps. Playwright for web, Electron for desktop, Maestro and Detox for mobile. Write a scenario and the AI harness pulls the test code along with it. Open the manual
blog-studio
A blog automation starting point. From research to drafting, review, and Blogger publish. An AI-smell gate and an AdSense gate right before publishing, plus simultaneous Korean–English publishing. Open the manual
book-studio
A book-writing starting point. A 4-Tier flow from Direction to Quality, a per-chapter AI-smell gate, a continuity index, and many books in parallel. Open the manual
scene-studio
A short-form video starting point. Two tracks, generative and remix, an approval gate at each stage, and a license policy planted in remix. Open the manual
legal-workflow
A legal-document processing harness for lawyers and legal teams. It runs locally on the lawyer’s PC so materials never leave it, with a lawyer review gate at each pipeline stage. Open the manual
What ddakit catches
Section titled “What ddakit catches”What each base handles ranges from a Chrome extension to writing a book, short-form video, and legal documents. Yet AI breaks in the same four places. The seven bases catch those four, each in its own way.
Caught on its own
book-studio won’t ship a chapter until it clears the AI-smell gate; if it doesn’t, the writing agent gets called again. blog-studio holds an AdSense policy gate right before publishing, so a post that would get ads rejected stops before it reaches Blogger. book-studio expert/verification · blog-studio expert/publishing
Back from a broken flow
chrome-ext-base prepends your last working spot to every new session, and /resume walks you back through a flow that dropped. saas-base points Claude Code and Codex at the same AI_AUTOMATION.md canonical, so decisions survive across sessions. chrome-ext beginner/when-ai-gets-stuck · saas-base expert/
Verification in one place
scene-studio puts an approval gate at the scene-breakdown, image, and video stages, so a person confirms before it moves on. legal-workflow holds a lawyer review gate at each pipeline stage, so nothing skips review on the way to the next document. scene-studio expert/tracks · legal-workflow
Licensee update channel
Shared across the forkable bases. /update-from-base applies the base-owned / user-owned / mixed classification so each new release flows into licensee forks. No more cherry-picking by hand. blog-studio expert/updates · book-studio expert/updates
How is this different from just asking AI?
Section titled “How is this different from just asking AI?”The start looks the same. Ask AI and the code, the prose, the video all come out plausible. The difference shows in whether the result survives to production, to publish, to filing.
Gates live in code, not in a doc
Other templates write “do it this way” in a README. ddakit’s gates actually stop the build, the publish, the commit. saas-base breaks the build the moment an FSD layer is crossed the wrong way, and chrome-ext-base turns a scenario line into the test code that confirms it. saas-base expert/architecture · chrome-ext expert/verification
Stuck twice in the same spot and it halts
Left alone, AI keeps going plausibly even when it’s stuck. chrome-ext-base catches the second stuck spot and /recover-from-blocked halts the run; book-studio hands off to you when AI-smell retries pass their limit. chrome-ext beginner/when-ai-gets-stuck · book-studio beginner/write-first-chapter
Non-developers get it with one command
A better base usually means copying changes over by hand. ddakit runs /update-from-base with owner-classified auto-merge even if you don’t know git, and one command ships Korean and English together. blog-studio beginner/bilingual · scene-studio expert/updates
Who this was built for
Section titled “Who this was built for”People who write code
AI gets you going fast, but does what you build survive to production? chrome-ext-base and saas-base pack same-mistake-twice auto-halt, scenario-driven checks, and build-time boundary enforcement into one license. chrome-ext manual · saas-base manual
People who write and make video
Where an AI tell costs you trust. blog-studio and book-studio run an AI-smell gate before publishing and drafting, and scene-studio puts an approval gate and a license policy at each stage. blog-studio manual · book-studio manual · scene-studio manual
People who handle sensitive documents
Where materials can’t leak. legal-workflow runs locally on the lawyer’s PC so nothing leaves it, with a lawyer review gate at each pipeline stage. legal-workflow manual
If you want the structure in depth, each base has its own expert track with architecture and verification. Start with saas-base architecture and book-studio verification.
PREVIEW
The manual is free. The base ships next as a license.
Section titled “The manual is free. The base ships next as a license.”This manual opens up what all seven bases cover, free. The base code and the .claude/ bundle (agents, skills, hooks) ship soon as a paid license, Solo single-seat plus a year of the update channel. The registration channel will be posted on this page as soon as it lands.
Category: Claude Code agentic boilerplate, AI guardrails for vibe coding.