agentleFS
bakhod1rbakhod1r/awesome-agents/.claude/skills/adr/SKILL.md

adr

Skill0 starsChanged 3 months agoMITView on GitHub

Write an Architecture Decision Record. Use when a decision is hard to reverse — technology choice, service boundary, data model, protocol, vendor, or anything a future engineer will ask "why on earth" about.

97 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use fdr, sweep and 8 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • thiagofgfSkill

    fdr

    Use when creating, changing, reviewing, retiring, auditing, or backfilling implemented user-visible feature behavior, permissions, feature-specific rationale, tradeoffs, or relationships to architectural decisions.

    Also in 3 projects that use this

  • drewdrewthisSkill

    sweep

    Search for similar patterns, anti-patterns, or issues elsewhere in the codebase after a fix or feature. Use after implementing a fix to catch the same bug in other places, or proactively to audit for a known pattern.

    Also in 4 projects that use this

  • vndeeSkill

    product-spec

    Use when starting a new feature, product, or project — before any architecture or code, to define what to build and why

    Also in 3 projects that use this

  • montymiSkill

    tech-spec

    Parse and review Technical Specification documents for completeness, consistency, and implementation readiness

    Also in 3 projects that use this

  • OpenAISkill

    observability

    Vercel Observability expert guidance — Drains (logs, traces, speed insights, web analytics), Web Analytics, Speed Insights, runtime logs, custom events, OpenTelemetry integration, and monitoring dashboards. Use when instrumenting, debugging, or optimizing application performance and user experience on Vercel.

    Also in 8 projects that use this

  • JohnJBorenSkill

    ci-cd

    CI/CD pipeline design, GitHub Actions, testing automation, and deployment strategies

    Also in 4 projects that use this

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 81 edited versions we read keep these.

  • 2

    Distinctive

    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
  1. Distinctive:

    Write an ADR when the decision is expensive to reverse.

    line 10
  2. Distinctive:

    Do not write one for: library version bumps, naming, formatting, or anything a single pull request can undo.

    line 13
  3. Very specific:

    Never delete an ADR — set Status: Superseded by NNNN.

    line 87

“Common” is measured: kept by at least 6 in 10 of the 81 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

ADR

When to write one

Write an ADR when the decision is expensive to reverse. The test: would undoing this in six months require a migration, a rewrite, or a contract negotiation? Then it needs a record.

Do not write one for: library version bumps, naming, formatting, or anything a single pull request can undo.

Location

docs/adr/NNNN-kebab-title.md. Sequential, zero-padded, never renumbered — links rot otherwise.

Template

# NNNN. <Title stated as the decision, not the problem>

- Status: Proposed | Accepted | Superseded by NNNN
- Date: YYYY-MM-DD
- Deciders: <names or agent roles>

## Context
<The forces. Facts and constraints only — no narrative, no history lesson.
What makes this hard? What did we measure? What are we not allowed to change?>

## Decision
<What we will do, active voice, present tense. "We will shard by tenant_id."
Not "it was decided that sharding might be considered.">

## Consequences
<What becomes easier. What becomes harder. What we now owe — the debt this creates.
Include the ugly ones; the ADR is worthless if it only lists benefits.>

## Alternatives Considered
<Each real option with the concrete, specific reason it lost.>

Worked example

Two versions of the same Alternatives section:

Weak — unfalsifiable, tells a future reader nothing:

## Alternatives Considered
- Sharding by user_id: not a good fit for our use case.
- Read replicas: didn't scale well enough.
- Bigger instance: too expensive.

Strong — every rejection is a checkable fact:

## Alternatives Considered
- **Shard by user_id.** Rejected: 38% of queries join across users within one tenant,
  which becomes cross-shard. Measured on 7 days of query logs (`analysis/shard-key-2026-03.md`).
- **Read replicas only.** Rejected: writes are the bottleneck, not reads.
  Write throughput is 4.2k/s against a measured single-primary ceiling of ~5k/s;
  this buys roughly four months of growth at the current 12%/month rate.
- **Vertical scale to the largest instance.** Rejected: +$18k/month for a 1.8x ceiling,
  and the next step after that does not exist. Delays the same decision by two quarters.

The second version survives the question "did you actually consider this?" The first does not.

Anti-patterns

Anti-patternFix
Consequences lists only benefitsEvery real decision costs something. Name it.
"Best practice" as a justificationCite the measurement or the constraint instead.
Alternatives that were never seriously consideredStraw men are worse than omission — delete them.
Deleting a superseded ADRSupersede it. The reasoning stays valuable even when wrong.
Written after implementation to satisfy processThe record exists to shape the decision, not to document it.
Two decisions in one ADRSplit. They will be superseded on different timelines.

Rules

  • One decision per record.
  • Never delete an ADR — set Status: Superseded by NNNN.
  • Quantify trade-offs in money, latency, or effort. "Faster" is not a consequence.
  • Deciders are named. An unowned decision is a rumour.

Done when

  • The title states the decision, not the problem.
  • Context contains facts and constraints — no story.
  • Consequences include at least one genuine cost.
  • Every alternative has a specific, checkable rejection reason.
  • A stranger could re-derive the decision from the record alone.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.