Architecture Decision Records
Published by spencerpauly in awesome-cursor-skills
What this skill does
Document technical decisions as Architecture Decision Records (ADRs) with context, options considered, and rationale.
Add Architecture Decision Records to your agent
Review the source and files first. When you are ready, copy the prompt instruction or use the CLI command supported by your environment.
Install with a prompt
Paste this into a compatible coding agent:
add this skill "architecture-decision-records" from https://github.com/spencerpauly/awesome-cursor-skillsInstall with the CLI
Run this command in a controlled environment after reviewing the repository:
npx skills add https://github.com/spencerpauly/awesome-cursor-skills --skill architecture-decision-recordsSkill instructions
Architecture Decision Records
Document significant technical decisions so future you (and your team) understands why choices were made.
When to Write an ADR
Write one when you're making a decision that:
- Is hard to reverse later
- Affects multiple parts of the system
- Involves tradeoffs between valid options
- Will be questioned by someone in 6 months
Examples: choosing a database, adopting a framework, changing auth strategy, restructuring the API, adding a build tool.
Template
Create files in docs/decisions/ (or adr/) with sequential numbering:
# ADR-001: Use PostgreSQL for primary database
## Status
Accepted | Proposed | Deprecated | Superseded by ADR-XXX
## Date
2026-04-10
## Context
What is the problem or situation that requires a decision?
Include constraints, requirements, and forces at play.
## Options Considered
### Option A: PostgreSQL
- Pros: ACID compliance, JSON support, mature ecosystem, free
- Cons: Requires managing connections, vertical scaling limits
### Option B: MongoDB
- Pros: Flexible schema, horizontal scaling
- Cons: No transactions across collections, eventual consistency issues
### Option C: PlanetScale (MySQL)
- Pros: Serverless, branching, managed
- Cons: Vendor lock-in, no foreign keys in their branching model
## Decision
We chose **PostgreSQL** because:
1. Our data is relational — users, teams, projects with clear relationships
2. We need ACID transactions for billing operations
3. JSON columns give us schema flexibility where needed
4. Prisma/Drizzle have excellent Postgres support
## Consequences
- We need to manage connection pooling (use PgBouncer or Prisma's built-in pool)
- Migrations must be backwards-compatible for zero-downtime deploys
- We accept vertical scaling limits and will shard later if needed
Workflow
- Identify the decision being made
- Research the options (at least 2-3 alternatives)
- Write the ADR using the template
- Review with the team (PR or discussion)
- Merge as accepted
- Reference from code comments when relevant:
// See ADR-001
File Naming
docs/decisions/
├── 001-use-postgresql.md
├── 002-adopt-trpc-over-rest.md
├── 003-switch-to-pnpm.md
└── template.md
Tips
- Keep ADRs short — 1-2 pages max
- Write them in the present tense ("We choose X" not "We chose X")
- It's OK to write an ADR after the fact if you forgot to write one during the decision
- Superseded ADRs should link to their replacement, not be deleted
- ADRs are not design docs — they capture the decision, not the full design
Files included
- SKILL.md

