replica-architect
Skill2 starsChanged 5 days agoMITView on GitHub
Plans the stack, database schema and API for an app clone, from the recon map replica-recon wrote. Picks boring, managed tech, turns the inferred data model into real SQL with indexes and access rules, lists every route by flow, and orders the build as a thin vertical slice first. Use when the user says "plan the clone", "what stack should I use", "design the database", "write the schema", "plan the API", "architecture for my version of X", or after /replica-recon finishes.
3 other projects use a version of this, 1 of them edited; 1 of 1 edited versions we compared changed its rules; 2 new projects in the last 4 weeks, against 0 the 4 before.
How people change it
1 of 1 edited versions changed its rules.
Of the 1 edited version we compared: 1 made changes of their own.
Rules people drop
- If there is no recon map, stop and run
/replica-reconfirst. Planning a clone from memory of what an app does is how you miss half of it.dropped by 1 of 1 - Use what the user already knows if they have a stack. Otherwise the default, because every part is managed, documented and cheap at zero users:dropped by 1 of 1
- Write each choice with one line of why. One database. No microservices. The clone does not need the original's architecture, it needs the original's features.dropped by 1 of 1
- Only official, public APIs with the user's own keys. Never the original app's private endpoints, even if they are visible in a browser.dropped by 1 of 1
- Write a line on each that applies: time zones and daylight saving, idempotency (webhooks arrive twice), race conditions, rate limits, file size limits, search, realtime, offline, email deliverability, multi-tenancy, GDPR deletion.dropped by 1 of 1
We split this file and each edited version into the rules they give an agent and match them by wording. A version that adds no rule of its own and has not changed since this file last did reads as an older text, so what it lacks is not counted as dropped. We keep no past texts of a file, so that is inferred, and an edited version that started from an older text still counts its gaps as drops.
Adopted by
Surging: 2 new projects took it in the last 4 weeks, none the 4 weeks before.thin data
Most recently taken up by
- dmminga1983/minga-replicaedited, first seen Nov 15, 2026
- realtahailyas/reverse-engineer-app-skillsword for word, first seen Oct 25, 2026
- Jakeschincariol/replica-skillword for word, first seen Oct 3, 2026
Other versions3 projects use a version of this.
Word for word · 2
- Jakeschincariol/replica-skill★ 399100% same
- realtahailyas/reverse-engineer-app-skills★ 3100% same
Edited · 1
- dmminga1983/minga-replica★ 118% same
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository108 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/composition-patterns/AGENTS.mdAGENTS.md
- skills/react-best-practices/AGENTS.mdAGENTS.md
- skills/react-view-transitions/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- skills/master-developer/references/CLAUDE.mdCLAUDE.md
- agent-fabricSkill
- api-anythingSkill
- creator-security-automationSkill
- feature-implementationSkill
- omniroute-free-llm-routingSkill
- repository-auditSkill
- security-reviewSkill
- unslopSkill
- youtube-studioSkill
- agent-developmentSkill
- agent-fabricSkill
- agent-security-hardeningSkill
- algorithmic-artSkill
- api-anythingSkill
- api-designSkill
- backend-patternsSkill
- brand-guidelinesSkill
- build-mcp-appSkill
- build-mcpbSkill
- build-mcp-serverSkill
- canvas-designSkill
- claude-apiSkill
- claude-md-improverSkill
- coding-standardsSkill
- command-developmentSkill
- vercel-composition-patternsSkill
- creator-security-automationSkill
- database-migrationsSkill
- deployment-patternsSkill
- dispatching-parallel-agentsSkill
- docker-patternsSkill
- e2e-testingSkill
- executing-plansSkill
- finishing-a-development-branchSkill
- frontend-a11ySkill
- frontend-designSkill
- frontend-patternsSkill
- hook-developmentSkill
- master-developerSkill
- mcp-builderSkill
- mcp-dependency-reviewSkill
- mcp-integrationSkill
- mdcp-self-checkSkill
- mdcp-token-meterSkill
- nextjs-turbopackSkill
- omniroute-free-llm-routingSkill
- plugin-settingsSkill
- plugin-structureSkill
- postgres-patternsSkill
- prisma-patternsSkill
- vercel-react-best-practicesSkill
- vercel-react-view-transitionsSkill
- receiving-code-reviewSkill
- replica-backendSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
If there is no recon map, stop and run /replica-recon first.
- Distinctive: line 23
Use what the user already knows if they have a stack.
- Distinctive: line 40
Write each choice with one line of why.
- Distinctive: line 92
Only official, public APIs with the user's own keys.
- Distinctive: line 97
Write a line on each that applies: time zones and daylight saving, idempotency (webhooks arrive twice), race conditions, rate limits, file …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
replica-architect
Reads replica/recon.md and replica/features.csv. Writes
replica/architecture.md (template: architecture.md in this folder).
If there is no recon map, stop and run /replica-recon first. Planning a
clone from memory of what an app does is how you miss half of it.
Step 1: the stack
Use what the user already knows if they have a stack. Otherwise the default, because every part is managed, documented and cheap at zero users:
| layer | default | swap for |
| web app | Next.js (App Router) + TypeScript | Remix, SvelteKit, Rails |
| styling | Tailwind, tokens from replica-design | CSS modules |
| mobile | Expo (React Native) | SwiftUI, Kotlin |
| database | Postgres on Supabase or Neon | PlanetScale, SQLite (Turso) |
| ORM | Drizzle or Prisma | raw SQL |
| auth | Supabase Auth or Auth.js | Clerk |
| payments | Stripe Checkout + Billing | Lemon Squeezy, Paddle |
| Resend or Postmark | SES | |
| jobs | Vercel Cron, Inngest or Trigger.dev | a worker on Fly |
| files | Supabase Storage or Cloudflare R2 | S3 |
| hosting | Vercel | Netlify, Fly, Render |
Write each choice with one line of why. One database. No microservices. The clone does not need the original's architecture, it needs the original's features.
Step 2: the schema
Turn the inferred data model into SQL. For every table:
id uuid primary key default gen_random_uuid(),created_at,updated_at- an owner column (
user_idororg_id) on everything a user owns - foreign keys with an
on deleterule decided, not defaulted - indexes on every foreign key and every column you filter or sort by
- enums or check constraints for status fields
- times as
timestamptz, always, stored in UTC - money as integer cents plus a currency column
- access rules: Postgres row level security on Supabase, or one authorisation check per query in the data layer. Write which.
Example, for a booking app:
create table bookings (
id uuid primary key default gen_random_uuid(),
event_type_id uuid not null references event_types(id) on delete cascade,
host_id uuid not null references users(id) on delete cascade,
start_at timestamptz not null,
end_at timestamptz not null,
guest_name text not null,
guest_email text not null,
guest_timezone text not null,
status text not null default 'confirmed'
check (status in ('confirmed','cancelled','rescheduled')),
answers jsonb not null default '{}',
created_at timestamptz not null default now(),
constraint no_zero_length check (end_at > start_at)
);
create index on bookings (host_id, start_at);Then the hard constraints the recon found. Two guests booking the same slot is a database problem (an exclusion constraint or a unique index), not a UI problem.
Step 3: the API
One table per flow from the recon map. For every route or server action:
method path | what it does | who can call it | input | output | flow
Plus webhooks in (Stripe, calendar providers) and out, and background jobs (reminders, sync, cleanup) with their schedule.
Only official, public APIs with the user's own keys. Never the original app's private endpoints, even if they are visible in a browser.
Step 4: the parts that bite
Write a line on each that applies: time zones and daylight saving, idempotency (webhooks arrive twice), race conditions, rate limits, file size limits, search, realtime, offline, email deliverability, multi-tenancy, GDPR deletion.
Step 5: build order
- Vertical slice. The core loop end to end, ugly: sign up, do the one thing, see the result. Proves the stack.
- Must-haves from
features.csv, by area. - Should-haves, then could-haves.
- The fixes replica-entrepreneur finds, once it has run.
Each milestone lists its screens (S-IDs), tables and routes.
Output
replica/architecture.md, the SQL in replica/schema.sql or as the first
migration, and a summary: stack in one line, table count, route count, the
three riskiest parts, and the next step: /replica-design.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

