agentleFS
Future-CXFuture-CX/AI-Architecture-Toolkit/skills/architecture-decision-record/SKILL.md

architecture-decision-record

Skill8 starsChanged 6 months agoMITView on GitHub

Record durable architecture decisions as concise ADRs using the repository ADR format. Use when a decision is hard to reverse, surprising without context, and the result of a real trade-off.

13 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. 0 are common, 9 are distinctive, and 8 are very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 9

    Distinctive

    This author's own choices. Worth a look.

  • 8

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when a design conversation produces a decision that should be remembered by future readers.

    line 10
  2. Very specific:

    When this toolkit is used as a submodule, do not write ADRs under toolkit/adr/.

    line 12
  3. Very specific:

    Start at adr-0001. For new ADRs, inspect existing adr/adr-*.md files and any legacy adr/0001-*.md files, choose the next available 4-digit …

    line 20
  4. Distinctive:

    When running from a private lab repo that contains this toolkit as a submodule:

    line 29
  5. Distinctive:

    Create ADRs only when all three conditions are true:

    line 47
  6. Distinctive:

    Write the opening summary as no more than two short paragraphs.

    line 57
  7. Very specific:

    Add a required Decision section immediately after the opening summary.

    line 58
  8. Very specific:

    Add a Recommendation section after Considered Options that names the preferred option and explains why it is stronger than the alternatives.

    line 61
  9. Very specific:

    Keep Consequences to 3-6 bullets when the downstream effects are worth calling out.

    line 62
  10. Distinctive:

    Avoid dense technical terms where a plain phrase works.

    line 63
  11. Distinctive:

    Run a final readability pass before publishing: shorten long sentences, split dense paragraphs, explain necessary terms, and remove …

    line 65
  12. Distinctive:

    Check whether the decision meets the ADR threshold.

    line 66
  13. Very specific:

    If it qualifies, write the ADR under the consuming repository's root adr/ folder using templates/adr-template.md and the …

    line 67
  14. Very specific:

    Add or update the ADR in adr/_adr-overview.md with a relative link, status with colored icon, summary, and last_updated date.

    line 68
  15. Distinctive:

    If it does not qualify, summarize the decision inline without creating an ADR.

    line 69
  16. Distinctive:

    Use ADR template for file location, numbering, structure, and optional sections.

    line 73
  17. Very specific:

    Use ADR overview template for the adr/_adr-overview.md table structure.

    line 75

“Common” is measured: kept by at least 6 in 10 of the 10 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

Architecture Decision Record

Quick Start

Use this skill when a design conversation produces a decision that should be remembered by future readers.

When this toolkit is used as a submodule, do not write ADRs under toolkit/adr/. Run the command from the private repository root, or pass --output-root <private-repo-root>, so output goes to the consuming repository.

ADR filenames must use this format:

adr-<4-digits>-<slug>.md

Start at adr-0001. For new ADRs, inspect existing adr/adr-*.md files and any legacy adr/0001-*.md files, choose the next available 4-digit number, and preserve gaps unless the user explicitly asks to renumber.

python3 skills/architecture-decision-record/scripts/write-adr.py "Use Domain Events Between Ordering And Billing" \
  --status accepted \
  --summary "Billing outages must not stop customers from completing checkout." \
  --decision "Ordering and Billing will communicate through domain events instead of synchronous HTTP."

When running from a private lab repo that contains this toolkit as a submodule:

python3 toolkit/skills/architecture-decision-record/scripts/write-adr.py "Use Domain Events Between Ordering And Billing" \
  --status accepted \
  --summary "Billing outages must not stop customers from completing checkout." \
  --decision "Ordering and Billing will communicate through domain events instead of synchronous HTTP."

Each generated ADR is also added to adr/_adr-overview.md using a markdown table with name, status, description, and last_updated columns. Create _adr-overview.md when it does not exist. Keep rows sorted alphabetically by ADR title.

Status values use colored icons in the overview:

  • proposed renders as 🟡 proposed
  • accepted renders as 🟢 accepted
  • deprecated renders as 🔴 deprecated
  • superseded renders as ⚫ superseded

Create ADRs only when all three conditions are true:

  1. The decision is hard to reverse.
  2. The decision is surprising without context.
  3. The decision is the result of a real trade-off.

Workflow

  1. Confirm the decision, context, and selected option.
  2. Identify the meaningful alternatives that were rejected.
  3. Write the opening summary as no more than two short paragraphs. Start with the business outcome, context, and operational impact.
  4. Add a required Decision section immediately after the opening summary. State the selected option and its important boundaries in direct language.
  5. Target a Readability Score of 40 or higher. If the score is below 40, simplify before publishing.
  6. When rejected alternatives are worth remembering, include a Considered Options section with the options listed as bullets first, followed by a comparison table. Use options as columns and include rows for Architecture Fit, Company Fit, Effort, and Complexity. Keep each table cell to one sentence, no more than about 15 words, and avoid implementation detail unless it changes the decision.
  7. Add a Recommendation section after Considered Options that names the preferred option and explains why it is stronger than the alternatives.
  8. Keep Consequences to 3-6 bullets when the downstream effects are worth calling out.
  9. Avoid dense technical terms where a plain phrase works. If a technical term is necessary, use it consistently and explain it in the first paragraph or in a glossary.
  10. End the ADR with a Reference Links section. List the Confluence page for the ADR Overview first and the Confluence page for the current ADR second. Replace the template placeholders with the actual Confluence URLs before finalizing the ADR.
  11. Run a final readability pass before publishing: shorten long sentences, split dense paragraphs, explain necessary terms, and remove implementation detail that does not affect the decision.
  12. Check whether the decision meets the ADR threshold.
  13. If it qualifies, write the ADR under the consuming repository's root adr/ folder using templates/adr-template.md and the adr-<4-digits>-<slug>.md filename format.
  14. Add or update the ADR in adr/_adr-overview.md with a relative link, status with colored icon, summary, and last_updated date.
  15. If it does not qualify, summarize the decision inline without creating an ADR.

Format

Use ADR template for file location, numbering, structure, and optional sections.

Use ADR overview template for the adr/_adr-overview.md table structure.

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.