agentsclimarketplace

Rfc

Skill yuri-semenenko/ai-engineering-workspace/claude-code/.claude/skills/rfc

Draft an architecture RFC using the user's canonical 10-section format (Problem → Context → Constraints → Assumptions → Options → Trade-offs → Recommendation → Risks → Migration Strategy → Open Questions). Use when the user asks for an "RFC", "design doc", "architectural proposal", or "RFC по <теме>". Optimized for the persona in ~/.claude/CLAUDE.md.From its SKILL.md

Install
npx -y skills add yuri-semenenko/ai-engineering-workspace --skill rfc

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.7 KB, 556 tokens by cl100k_base, as published. Nobody here has run it

RFC

Produce an architecture RFC in the user's canonical format. The user is a Staff Engineer biased toward simplicity, FP, modular monoliths, and pragmatic DDD — challenge over-engineering inside the doc itself, do not defer it to review.

Steps

  1. Clarify scope before writing. If the input is vague ("RFC for caching"), ask 1-3 targeted questions: what problem prompted it, what's the current state, what's the deadline / urgency. Do not invent context.
  2. Read the repo if relevant — package.json, top-level structure, existing ADRs/RFCs (docs/rfc/, docs/adr/, .docs/) to mirror naming and tone.
  3. Draft all 10 sections. Do not skip any. If a section is genuinely empty, write N/A — <one-line reason> rather than removing it.
  4. Surface 2-3 realistic options. Never present a single recommendation as "the" answer. Include a "do nothing / defer" option when honest.
  5. Trade-offs must be concrete. Cost in dev hours, operational complexity, blast radius if it fails. No marketing prose.
  6. Recommendation must justify itself against the other options, not in isolation.
  7. Risks ≠ trade-offs. Risks are what could go wrong despite the right choice (regression, scope creep, dependency change, hiring gap).

Required sections (in order)

  1. Problem Statement — one paragraph, no jargon. What hurts today.
  2. Context — current architecture, team, constraints from history. Link to relevant code/docs.
  3. Constraints — hard limits (compliance, deadline, headcount, infra).
  4. Assumptions — explicit, refutable. Each one should be a sentence someone could disagree with.
  5. Options — 2-4 alternatives. Include status quo.
  6. Trade-offs — table preferred: option × (dev cost, ops cost, time-to-value, reversibility, risk).
  7. Recommendation — chosen option + why it beats each alternative.
  8. Risks — what we accept by choosing this. Include mitigations only where they're real.
  9. Migration Strategy — phased plan if non-trivial. Define rollback criteria.
  10. Open Questions — things this RFC explicitly does not answer.

Output

Single markdown document. Use H2 (##) for each of the 10 sections. Tables for trade-offs. No emoji. No "this comprehensive proposal aims to leverage" prose — concise, structured, direct.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,736. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.