adr-author
Skill10 starsChanged 3 months agoMITView on GitHub
Write a new Architecture Decision Record following the established Nygard format and conventions.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- rajanrx/skillstheir own version, first seen Nov 8, 2026
- jaskirat1616/grok-skillstheir own version, first seen Nov 1, 2026
- JayRHa/agent-skillstheir own version, first seen Oct 15, 2026
- microsoft/hve-coretheir own version, first seen Oct 13, 2026
- Atypical-Consulting/AdrMcptheir own version, first seen Jul 8, 2026
- JayRHa/AgentSkillstheir own version, first seen Jun 9, 2026
Used with
Projects with it also use 0-template, github-actions-builder and 10 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.
JayRHaSkill0-template
Template for creating a new Agent Skill. Copy this folder, rename it to your skill's kebab-case name, and fill in every section. Use this as the starting point whenever you author a new skill for the AgentSkills library.
Also in 3 projects that use this
JayRHaSkillgithub-actions-builder
Builds reliable, fast, and secure GitHub Actions CI/CD workflows with dependency caching, build matrices, scoped secrets, concurrency control, and reusable/composite workflows. Use this skill when the user asks to "create a GitHub Actions workflow", "set up CI", "add CI/CD", "fix a flaky/slow GitHub Actions pipeline", "add a build matrix", "cache dependencies in CI", "write a reusable workflow", "configure deployment with environments and secrets", or mentions files under .github/workflows.
Also in 3 projects that use this
xmqywxSkillmeeting-summarizer
Transform meeting transcripts into structured summaries with action items, decisions, follow-ups, and efficiency analysis. Works with any meeting format.
Also in 3 projects that use this
MehdibargachSkillokr-writer
Turn business goals into structured OKRs with measurable Key Results, linked initiatives, alignment check, and anti-goals. Takes a strategic priority and outputs ready-to-use OKRs.
Also in 3 projects that use this
pbc-osSkillsecrets-manager
Set up and configure cloud secrets management for your AI agent. Guides users through platform selection (GCP, AWS, Azure, 1Password, Doppler, HashiCorp Vault), account setup, CLI configuration, and secure credential storage.
Also in 3 projects that use this
- +7 more from JayRHa
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- microsoft/hve-core★ 1.6ksame name
- jaskirat1616/grok-skills★ 22same name
- JayRHa/agent-skills★ 12same name
- JayRHa/AgentSkills★ 12same name
- rajanrx/skills★ 2same name
- Atypical-Consulting/AdrMcp★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- instructions/claude.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- analyze-r-packageSkill
- bioc-howtoSkill
- bioc-pkg-devSkill
- bioc-pkg-finderSkill
- bioc-pkg-reviewSkill
- build-check-bioccheckSkill
- check-bioconductor-skillsSkill
- create-package-instructionsSkill
- create-skillSkill
- document-skillSkill
- improve-code-coverageSkill
- migrate-datapkg-to-zenodoSkill
- security-audit-r-packageSkill
- update-package-instructionsSkill
- update-r-newsSkill
- validate-skillSkill
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 6 are distinctive, and 5 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
6
Distinctive
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Write a new Architecture Decision Record (ADR) following the established Nygard format and conventions.
- Very specific: line 37
Supersession: Never rewrite an accepted ADR's Context, Decision, or Consequences to say something it did not say.
- Very specific: line 49
Index: Add a line to the index file, e.g., docs/adr/README.md.
- Distinctive: line 54
Do NOT write ADRs for routine bug fixes, version bumps, or anything self-evident from the diff.
- Distinctive: line 62
Before bundling, ask of each decision in the draft:
- Distinctive: line 64
Could it be reversed on its own?: If yes, and it passes the warranted test above on its own, it needs its own record.
- Distinctive: line 65
Would someone search for it by name?: If the decision is not findable from the ADR's title, it is in the wrong ADR.
- Very specific: line 85
Read the 2-3 latest ADRs in the repository (e.g., in docs/adr/) to understand the established tone and context.
- Very specific: line 89
If the new ADR supersedes an older one, edit the old ADR's Status to Superseded by NNNN.
- Very specific: line 90
Add an entry for the new ADR to the index file (docs/adr/README.md).
- Distinctive: line 91
Ask the user for explicit approval before committing any changes.
“Common” is measured: kept by at least 6 in 10 of the 5 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.
Files
Authoring an ADR
Write a new Architecture Decision Record (ADR) following the established Nygard format and conventions. Use this skill when a non-trivial, behavior-affecting decision is made and should be recorded, or the user says "write an ADR", "record this decision", or "document the design rationale".
Usage
Invoke this skill when you want to create a new Architecture Decision Record.
- "Write an ADR for..."
- "Record the decision to..."
- "Document the design rationale for..."
Prerequisites
- The repository keeps ADRs in a standard location, typically
docs/adr/. - An index of ADRs (e.g.,
docs/adr/README.md) and a template (e.g.,docs/adr/template.md) usually exist in this directory. If atemplate.mddoes not exist in the target repository, copy thetemplate.mdprovided alongside this skill into the target repository'sdocs/adr/directory so it is available for future use.
Conventions
- Filename:
NNNN-kebab-title.md, zero-padded sequential number. The next number should be the highest existing number plus 1 (e.g., checkls docs/adr/). - Format: Sections must include
Context→Decision→Alternatives considered→Consequences, preceded by a metadata block. Nygard's original elements are Title, Context, Decision, Status and Consequences;Alternatives consideredis a later and near-universal addition, kept here because it is the part future readers actually reread.# NNNN. <Title in sentence case> - **Status:** Accepted (or: Proposed | Superseded by NNNN) - **Date:** YYYY-MM-DD - **Deciders:** <Names> Supersession: Never rewrite an accepted ADR's Context, Decision, or Consequences to say something it did not say. Write a NEW ADR and mark the old one
Superseded by NNNN; the new ADR sayssupersedes MMMM.What is expected to change on the old record is its status and its forward reference — Nygard's own words are "if a decision is reversed, we will keep the old one around, but mark it as superseded", "with a reference to its replacement". So editing an accepted ADR to point at its replacement is the mechanism, not an exception to it. A short banner under the metadata block naming what changed, and leaving the original prose untouched below it, is preferred to a bare status line: a reader who lands on the old record mid-search needs to know what is still in force before reading on.
Partial correction:
deprecatedandsupersededboth retire a whole record, so neither is true when a later ADR changes part of an earlier one that otherwise stands. Do not mark such a record superseded. Instead, name the amended part in its Status line and add a banner directly under the metadata block saying what changed and what still stands:- **Status:** Accepted; decision §1 amended YYYY-MM-DD > **Amendment (YYYY-MM-DD):** decision §1 below places X in this repository. It has since moved to > Y, because <reason>. The rest stands unchanged: <what survives>.- Index: Add a line to the index file, e.g.,
docs/adr/README.md.
When an ADR is Warranted
Record decisions that change runtime behavior, are expensive to reverse, or are non-obvious from the code (e.g., retry/error policies, storage/publish layouts, framework choices, container strategies). Do NOT write ADRs for routine bug fixes, version bumps, or anything self-evident from the diff.
Scoping an ADR
The "warranted" test above decides whether to write a record. This decides how many.
One decision per record. The test is not length or topic breadth, it is supersession granularity: an ADR's value is that it can be pointed at, and reversed, on its own. Bundle two decisions and you have coupled their futures — reversing one means either superseding a record that also holds the other, or bolting on an amendment banner that every later reader has to parse before reaching the decision they came for.
Before bundling, ask of each decision in the draft:
- Could it be reversed on its own? If yes, and it passes the warranted test above on its own, it needs its own record. A reversible choice that was never architectural does not become one by being separable.
- Would someone search for it by name? If the decision is not findable from the ADR's title, it is in the wrong ADR. A decision about what a status value means does not belong in a record titled for validation rules.
- Does it share a context with the others, or just a sitting? Decisions made in the same afternoon are not thereby one decision. A shared triggering incident is cohesion; a shared calendar slot is not.
Split by weight, not only by count. A draft that bundles four decisions is usually not four ADRs — more often two of them never warranted a record at all. Re-apply the warranted test to each item separately:
- A field's optionality, a naming convention, an error-message format → belongs in the specification or the code it describes, not in the decision log.
- An implementation choice nothing outside the module can observe → belongs in a code comment and the commit message.
- A choice that changes what conforming input looks like, or that a future maintainer could plausibly reverse → its own ADR.
A decision log cluttered with small records is a real cost, but the usual cause is recording things that were never architectural, not splitting things that were.
Content Guidance
- Context = The forces and the triggering experience, described concretely. Real incidents and specifics beat abstractions.
- Decision = What will be done, including actual config or code snippets if relevant.
- Alternatives considered = Each rejected option AND why it was rejected. This is the part future readers actually reread.
- Consequences = What gets better, and what new costs or trade-offs are accepted.
Process
- Read the 2-3 latest ADRs in the repository (e.g., in
docs/adr/) to understand the established tone and context. - Determine the next sequential number
NNNNfor the new ADR. - Copy the
template.md(from the repository or from this skill's directory) todocs/adr/NNNN-<kebab-case-title>.md. If the target repository did not have a template, also save a copy of the template todocs/adr/template.md. - Fill in the template following the content guidance above.
- If the new ADR supersedes an older one, edit the old ADR's Status to
Superseded by NNNN. - Add an entry for the new ADR to the index file (
docs/adr/README.md). - Ask the user for explicit approval before committing any changes. If approved, commit the changes with the related code change (or alone if it's purely a decision).
Examples
Example: Authoring an ADR
User: "Write an ADR for switching our retry policy to exponential backoff." Agent:
- Checks
docs/adr/for the latest ADR to get the tone and next sequence number. - Copies
docs/adr/template.mdtodocs/adr/0012-exponential-backoff-retry-policy.md. - Fills out the Context, Decision, Alternatives Considered, and Consequences.
- Adds the new ADR to
docs/adr/README.md. - Prompts the user to review the ADR.
Notes
- See
template.mdin this skill directory for a reference Nygard ADR template.
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
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.

