agentsclimarketplace

Architecture decision records

Skill Amey-Thakur/AI-SKILLS/skills/architecture/architecture-decision-records

Capture each significant architecture decision as a short, immutable record of context, decision, and consequences with a tracked status. Use when a choice is costly to reverse or crosses teams and the reasoning must outlive the people who made it.From its SKILL.md

Install
npx -y skills add Amey-Thakur/AI-SKILLS --skill architecture-decision-records

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

2 things to look at

  • 23 days oldThe repository was created 23 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 4 stars4 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, 574 tokens by cl100k_base, as published. Nobody here has run it

Architecture decision records

A decision made in a meeting evaporates the moment the room clears, and six months later someone reverses it without knowing what it cost the last time. An ADR is the durable answer to "why is it built this way": one file, written once, that a future maintainer can trust more than anyone's memory.

Method

  1. One decision per record, numbered and immutable. Name files docs/adr/0007-use-postgres-for-billing.md. The number never changes and the file is append-only after acceptance. A record that mixes two decisions cannot be superseded cleanly when only one of them ages out.
  2. Write the context as forces, not narrative. State the constraints, requirements, and pressures in play: throughput targets, team skills, existing contracts, deadlines. The reader must feel the pressure that made the decision reasonable at the time, even if it looks wrong later.
  3. State the decision in the active voice. "We will store billing events in Postgres and stream them to the warehouse nightly." Not "Postgres seems like a good option." A decision hedged into a suggestion is not a decision.
  4. Record consequences honestly, both signs. List what gets easier and what gets harder: the new operational burden, the capability you gave up, the follow-on work this forces. An ADR with only upsides is marketing.
  5. Track status through a fixed lifecycle. Proposed, accepted, deprecated, superseded. When a later ADR overrides this one, mark it "Superseded by ADR-0021" and link both directions. Never edit the decision of an accepted record: write a new one that replaces it.
  6. Keep the record next to the code it governs. Store ADRs in the repo, not a wiki that drifts. Tools like adr-tools or the MADR template scaffold the file and the index; a link from the affected module's README closes the loop.

Litmus tests

  • Can a new engineer read the record and defend the decision to a skeptic without asking anyone what really happened?
  • Does every accepted ADR name at least one consequence the team dislikes?
  • When a decision is reversed, can you find the superseding record from the old one in a single click?

Boundaries

Record decisions that are expensive to reverse or affect more than one team: routine, two-way-door choices belong in the commit message, not an ADR. Adopt your organization's existing RFC or ADR template where one is in use rather than inventing a parallel format.

What ships with it

Read from the repository

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

Keep looking

Skills are one crate of 326,790. 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.