/reference-app

Point Claude at another fully-developed application the user wants to use as a reference. Build a structured understanding of it. Store that understanding in memory so future sessions can compare the current project against it.

When to use

• The user is starting a new project and wants it modeled on one they've already built.
• The user is reviewing the current project's code and wants you to compare it against an existing well-built example.
• The user is migrating, refactoring, or modernizing and the reference app shows the destination state.
• The user wants you to "understand" their full app even though you're working in a different directory today.

Operating method — ASK FIRST, THEN ANALYZE

Step 0: Confirm intent.

Before reading anything, ask:

• "What would you like me to learn from this reference app?" — pattern conventions / architecture / specific module / all of it.
• "What does the current project share with the reference?" — same domain / same stack / similar feature / different / unsure.
• "Should I treat the reference as authoritative (apply its patterns here) or as inspirational (consider its patterns)?"

These three answers shape the depth and tone of the resulting memory. Don't skip them.

Step 1: Validate the path.

• Confirm the path exists and is readable.
• Check it has a recognizable project structure (package.json, pyproject.toml, go.mod, Cargo.toml, etc.). If not, ask the user what kind of project it is.

Step 2: High-level scan (light pass).

Read in this order, one file each, no deep dives yet:

• Top-level README.
• CLAUDE.md if it exists.
• The package/dependency manifest.
• The deployment / CI config.
• The directory tree (one level deep, maybe two).

Step 3: Architecture map.

• Identify the entry points (server start, client root, CLI commands).
• Identify the main modules and their responsibilities (one sentence each).
• Identify external dependencies that matter to the architecture (DB, queue, cache, auth, payment, etc.).
• Identify the testing approach (framework, layering).
• Identify the deploy target.

Step 4: Convention mining.

Read 5–10 representative files (one component, one route, one test, one util, one type definition, one style file). Extract:

• Naming conventions (camelCase, kebab-case, file-per-export, etc.).
• File organization (feature folders vs technical folders, co-located tests).
• Error handling pattern.
• API contract style (REST/GraphQL, error shape).
• State management approach.
• Styling approach.
• Logging and observability.
• Auth and session pattern.

Step 5: Distinctive choices.

What did this project do that's different from defaults? These are the high-information signals — they reveal taste and constraints.

• Custom build tooling.
• Hand-rolled abstractions where the framework would have done it.
• Strong opinions reflected in lints, hooks, or architecture rules.
• Specific tradeoffs visible in code comments or commit messages.

Step 6: Write the memory.

Create .claude/memory/reference/<reference-app-name>.md with the schema:

---
name: ref-<app-name>
type: reference
description: <one-line, e.g., "Reference app: AcmeShop — Next.js + Drizzle + Postgres SaaS, used as architecture model">
created: <YYYY-MM-DD>
last-verified: <YYYY-MM-DD>
source-path: <path>
authority-level: authoritative | suggestive | inspirational
---

# Reference: <app-name>

> Authority: <as set by user in step 0>. Cite when applying.

## Stack
- Language(s):
- Frameworks:
- Database:
- Auth:
- Deployment:
- Testing:

## Architecture (one paragraph + a small Mermaid diagram if it helps)

## Key conventions
- File organization:
- Naming:
- API style:
- State management:
- Error handling:
- Styling:
- Tests:

## Distinctive choices
- Choice 1: what they did, what it cost, what it bought.
- Choice 2: …

## Useful prior art for the current project
(populated when comparing to current — left blank initially)

## What NOT to copy
(blank initially; filled in when patterns are noted to be tied to the source's specific constraints)

Add an entry to .claude/memory/MEMORY.md under the Reference section.

Step 7: Confirm with the user.

Show the resulting memory. Ask:

• "Does this match your understanding of the reference?"
• "Anything to correct or add?"
• "What part of this should I apply when working on the current project?"

When the user later asks for a code review or build help

The memory is now in context. Use it like this:

• When the current project does something differently from the reference: surface it and ask. "Reference uses Drizzle; current project is using raw SQL queries. Intentional, or should we standardize?"
• When suggesting an approach: cite the reference. "In acmeshop (your reference) you used a feature-folder layout — same here?"
• Don't silently import a pattern. Always ask: "Reference handles this with X; want the same here?"

Boundaries

• Never modify the reference app — read-only.
• Never assume the reference is current best practice. Patterns from one project may be wrong for another.
• Never bury the source. Every recommendation drawn from the reference cites it.
• Don't analyze without permission. Step 0 is mandatory.
• If the reference contains secrets visible in committed files, redact them in memory and tell the user they're in the reference repo (so they can rotate).

Output format

## Reference: <app-name>
- Path: <path>
- Authority: <suggestive | authoritative>
- Memory written: .claude/memory/reference/<app-name>.md

## Quick summary
- Stack: <one line>
- Architecture shape: <one line>
- 3 most distinctive choices:
  1. …
  2. …
  3. …

## Ready to use
You can now ask things like:
- "Compare the auth in the current project to <ref>."
- "How does <ref> handle billing? Set up the same here."
- "Where does <ref> put feature flags? Should we follow that?"
- "Review my current code against <ref>'s conventions."