rfc

---

name: rfc description: Create Requests for Comments (RFCs) — proposals in docs/decisions/ opening a significant decision for discussion before committing. Use when the user wants to propose a change, draft a technical proposal, or open a design discussion. PROACTIVELY suggest for migrations, breaking API changes, new architecture, or competing approaches needing structured input. Triggers: "RFC", "propose", "design proposal", "request for comments", "get feedback on this approach", "I want input from the team". For decisions already made use adr. Not for small choices, general docs, or post-mortems. allowed-tools: Bash(git log *), Bash(git diff *)

Requests for Comments (RFCs)

Create structured proposals that open a significant decision for discussion before the team commits to an approach. RFCs live in the repo alongside ADRs so the full decision lifecycle — proposal, debate, outcome — is captured in version control.

An RFC answers: "Here's a problem, here's what I think we should do about it, and here's what I want your input on."

When the RFC concludes, it should be converted to an ADR (use /adr) to record the final decision. Both documents are kept — the RFC preserves the reasoning and alternatives explored; the ADR is the authoritative record.

When to suggest an RFC

An RFC is the right tool when a decision is significant enough to benefit from structured input before committing. Signals:

• Multiple valid approaches exist and trade-offs aren't obvious
• The change affects consumers, other teams, or system boundaries
• A migration or breaking change is being proposed
• The author wants buy-in before investing in implementation
• Someone says "I'm not sure which way to go" on a non-trivial question

If consensus already exists or the decision is small, an ADR (without a preceding RFC) is usually sufficient.

Directory structure and numbering

RFCs live in docs/decisions/ alongside ADRs:

docs/decisions/
  INDEX.md
  RFC-001-auth-strategy.md
  RFC-002-api-versioning.md
  ADR-001-use-react-hooks.md

File naming: RFC-{NNN}-{slug}.md — zero-padded 3-digit ID, kebab-case slug from the title.

Auto-incrementing: Scan docs/decisions/ for existing RFC-*.md files, find the highest number, use the next one. If the directory doesn't exist, create it and start at 001.

RFC template

---
id: RFC-{NNN}
title: "{Title}"
status: draft | open | closed | withdrawn
date: {YYYY-MM-DD}
author: {name}
related: []
tags: []
---

# RFC-{NNN}: {Title}

## Summary

A one-paragraph elevator pitch: what is being proposed and why it matters.
Someone skimming the index should be able to decide whether to read further
based on this summary alone.

## Problem statement

What problem does this solve? Why does it need solving now? Describe the
forces at play — technical limitations, user pain points, scaling concerns,
business requirements. Include enough context that someone unfamiliar with
the codebase understands the motivation without reading source code.

## Goals and non-goals

**Goals:**
- What this proposal aims to achieve

**Non-goals:**
- What is explicitly out of scope — this prevents scope creep and
  clarifies boundaries for reviewers

## Proposed solution

Describe the recommended approach. Be specific about the choices and
trade-offs that matter for the decision, but don't write a full design
spec — focus on *why this approach* over the alternatives.

Include enough technical detail that reviewers can evaluate feasibility,
but defer implementation minutiae to the actual work.

## Alternatives considered

### {Alternative A}

What it is, why it was considered, and what made it less attractive than
the proposed solution. Be fair — if an alternative is genuinely close,
say so.

### {Alternative B}

Same treatment. The goal is to show that the design space was explored,
not just to justify a foregone conclusion.

## Risks and trade-offs

What could go wrong? What are we giving up? Be honest — reviewers trust
RFCs that name their own weaknesses. Include mitigation strategies where
possible.

## Rollout plan

How would this be implemented and adopted? Consider:
- Can it be done incrementally or does it require a big-bang migration?
- What's the rough sequence of work?
- Are there dependencies or prerequisites?
- How will we know it's working? (metrics, tests, user feedback)

This section can be brief for early-stage proposals and expanded as the
RFC matures.

## Open questions

- [ ] {Question that needs input from reviewers}
- [ ] {Unresolved technical detail}
- [ ] {Organizational or process question}

## Resolution

_To be filled in when this RFC is closed._

**Decision:** {what was decided}
**Date:** {when}
**ADR:** [ADR-{NNN}](ADR-{NNN}-slug.md)

Section guidance

Section	Required?	Notes
Summary	Yes	Keep it to one paragraph
Problem statement	Yes	The foundation — if this isn't clear, nothing else matters
Goals and non-goals	Recommended	Especially valuable for large or ambiguous proposals
Proposed solution	Yes	The core of the RFC
Alternatives considered	Yes	At least one alternative, even if it's "do nothing"
Risks and trade-offs	Recommended	Builds trust, surfaces concerns early
Rollout plan	Optional	More important for migrations and breaking changes
Open questions	Yes	Even if just one — signals this is genuinely open for input
Resolution	Filled later	Left blank until the RFC concludes

RFC lifecycle

draft → open → closed (decision made) → ADR created
                  └→ withdrawn (abandoned, no decision)

• Draft: Author is still writing, not ready for review.
• Open: Published for discussion. Update the status field and share with reviewers.
• Closed: A decision has been reached. Fill in the Resolution section and create an ADR via /adr.
• Withdrawn: The proposal was abandoned — document why briefly in the Resolution section so future readers know it was a deliberate choice, not an oversight.

Converting an RFC to an ADR

When an RFC reaches a conclusion:

1. Update the RFC's status to closed
2. Fill in the Resolution section with the decision, date, and link to the new ADR
3. Create a new ADR (use /adr) — pull context and alternatives from the RFC rather than rewriting from scratch
4. Cross-reference: the ADR's related field includes the RFC ID; the RFC's resolution links to the ADR
5. Keep both documents — the RFC preserves discussion history; the ADR is the authoritative outcome

Index management

After creating or modifying an RFC, update the RFCs table in docs/decisions/INDEX.md. If the index doesn't exist, create it. Sort newest first.

Writing guidance

An RFC is an invitation to think together, not a decree. Write it to be challenged.

• Frame the problem before the solution. If reviewers don't agree there's a problem, they won't engage with the proposal constructively.
• Steel-man the alternatives. Presenting weak alternatives makes the RFC look like a rubber stamp. If an alternative is genuinely close, say so — it invites useful discussion about the margin.
• Name what you don't know. Open questions aren't a sign of weakness. They're the whole point — you're asking for input because the decision is genuinely hard.
• Keep the scope proportional. A focused RFC with clear boundaries gets better feedback than an expansive one that tries to solve everything at once.
• Write for the skeptical reviewer. Assume the reader's first instinct is "why can't we just keep doing what we're doing?" Answer that in the problem statement.