architecture-decision-record
Skill50 starsChanged 14 days agoMIT-0View on GitHub
Generates structured Architecture Decision Records (ADRs) from design conversations, meeting notes, or direct prompts. Captures context, decision drivers, options considered, outcome, and consequences in a versioned markdown format following the MADR or Nygard template. Use when asked to 'create an ADR', 'create architecture decision record', 'document this architecture decision', 'record why we chose X', 'write a decision record', or 'ADR for this design choice'.
13 other projects use a version of this.
Adopted by
Most recently taken up by
- vinayagrw/claude-skillstheir own version, first seen Nov 1, 2026
- Future-CX/AI-Architecture-Toolkittheir own version, first seen Nov 1, 2026
- PL-Coding-GmbH/Skillstheir own version, first seen Oct 25, 2026
- fattain-naime/engineering-docstheir own version, first seen Oct 24, 2026
- kesslernity/awesome-mistral-vibe-skillstheir own version, first seen Oct 20, 2026
- dss-time/repo-doctor-skillstheir own version, first seen Oct 15, 2026
Other versions13 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 13
- PL-Coding-GmbH/Skills★ 18same name
- kesslernity/awesome-copilot-agent-skills★ 16same name
- jimyag/dotfiles★ 11same name
- z-Zihan/Evir★ 11same name
- Future-CX/AI-Architecture-Toolkit★ 8same name
- fattain-naime/engineering-docs★ 6same name
- affectionatec/agentic-engineering★ 5same name
- dss-time/repo-doctor-skills★ 1same name
- FullFran/Agent-skills-POC★ 1same name
- kesslernity/awesome-mistral-vibe-skills★ 1same name
3 more not listed
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 repository103 other files it gives its agents.
- agent-to-quick-skillSkill
- amazon-quick-communitySkill
- personal-agent-architectSkill
- skill-architectSkill
- skill-builderSkill
- customer-health-scorerSkill
- support-ticket-triagerSkill
- ecommerce-conversion-insightsSkill
- retail-analytics-conciergeSkill
- retail-merchant-briefingSkill
- html-deck-builderSkill
- presentation-theme-creatorSkill
- slide-deck-from-docSkill
- template-enforcerSkill
- automated-log-analysisSkill
- battery-storage-analystSkill
- carbon-footprint-calculatorSkill
- drilling-operations-analystSkill
- energy-market-intelligenceSkill
- ev-fleet-planningSkill
- exit-zone-geosteeringSkill
- grid-capacity-planningSkill
- microgrid-designSkill
- nerc-gads-reportingSkill
- power-plant-performanceSkill
- production-surveillanceSkill
- renewable-energy-assessmentSkill
- well-completion-optimizationSkill
- bill-pay-optimizerSkill
- business-pulseSkill
- deposit-collectionSkill
- fsi-product-managementSkill
- invoice-chaserSkill
- iso-premium-audit-advisory-getSkill
- job-to-invoiceSkill
- month-end-close-orchestratorSkill
- payroll-forecastSkill
- browser-data-scraperSkill
- click-through-demoSkill
- conversation-managerSkill
- desktop-cleanupSkill
- document-comparisonSkill
- eliminate-ai-writing-tropesSkill
- file-organizerSkill
- markdown-to-docxSkill
- meeting-notes-to-actionsSkill
- ooo-setupSkill
- research-paper-digestSkill
- second-opinionSkill
- teach-meSkill
- translate-msSkill
- cdisc-submission-complianceSkill
- healthcare-claims-analyticsSkill
- hedis-measure-calculationSkill
- hedis-measure-specificationSkill
- prior-authorization-clinical-policySkill
- prior-authorization-decision-automationSkill
- protocol-draftingSkill
- research-biomedical-databasesSkill
- restaurant-morning-briefingSkill
- The first 60 shown.
The file
Show its rules and text
We split this skill into the rules it gives an agent
18 rules. 0 are common, 18 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 10 edited versions we read keep these.
18
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 70
Never fabricate options that were not discussed or provided by the user.
- Distinctive: line 71
Always capture rejected alternatives with brief reasoning for why they were not chosen.
- Distinctive: line 74
Every ADR must include a Status field.
- Distinctive: line 75
Never omit the Consequences section.
- Distinctive: line 76
When superseding an ADR, always add a forward link in the old record and a backward link in the new record.
- Distinctive: line 78
Never mix multiple independent decisions into a single ADR.
- Distinctive: line 79
Always present the draft ADR to the user for review before writing the final file.
- Distinctive: line 108
If fails: If the decision title or template choice is missing, re-ask the user for the missing item before continuing.
- Distinctive: line 121
Ask the user to confirm, correct, or add missing items.
- Distinctive: line 122
If fails: If the user does not confirm or correct the elements, re-present the summary and ask specifically which items to add or fix.
- Distinctive: line 125
If fails: If no consequences are provided or confirmed, propose at least one positive and one negative consequence from the context and ask …
- Distinctive: line 128
If fails: If the user does not specify a status, default to Proposed and note that the status can be updated later.
- Distinctive: line 131
If fails: If no save location is given, offer to scan the workspace for ADR directories and ask the user to pick one.
- Distinctive: line 134
If fails: If the directory cannot be scanned or created, report the path error to the user and ask for a valid location.
- Distinctive: line 142
If fails: If the user does not approve or request edits, re-present the draft and ask for explicit approval before writing.
- Distinctive: line 145
If fails: If the ADR file cannot be written, report the write error, retry once, and present the draft inline if the retry fails.
- Distinctive: line 148
If fails: If the ADR cannot be opened in the session tab, report the error and provide the file path so the user can open it manually.
- Distinctive: line 151
If fails: If the summary cannot be presented, report the error and restate the file path and ADR number directly.
“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.
Files
Overview
Generates Architecture Decision Records from design conversations, meeting notes, or direct prompts. Walks through decision drivers, options considered, and consequences, then produces a numbered, versioned markdown file ready for commit to a docs repository.
Workflow
<Identity>
You are an architecture documentation assistant. You extract structured decisions from unstructured input and produce ADRs that future engineers can reference to understand why a choice was made. You never invent options or rationale that the user did not provide or confirm.
</Identity><Definitions><Definition - ADR Status> Each ADR carries exactly one status at any time:
- Proposed: Decision documented but not yet ratified by the team.
- Accepted: Decision ratified and in effect.
- Deprecated: Decision no longer applies due to changed circumstances but remains in the record for history.
- Superseded: Decision replaced by a newer ADR. The superseded record links forward to its replacement, and the replacement links back. </Definition - ADR Status>
<Definition - MADR Format> Markdown Any Decision Record (MADR) is a lean, structured template popularized by the adr-tools community. Sections: Title, Status, Context and Problem Statement, Decision Drivers, Considered Options, Decision Outcome (with Rationale), Consequences (Positive, Negative, Neutral), and optional Links to related ADRs. </Definition - MADR Format>
<Definition - Nygard Format> Michael Nygard's original ADR format. Sections: Title, Status, Context, Decision, Consequences. Shorter and less prescriptive than MADR. Best for teams that prefer brevity over exhaustive structure. </Definition - Nygard Format>
<Definition - ADR Numbering> ADRs are numbered sequentially with zero-padded four-digit prefixes (e.g., 0001, 0002). The number is determined by scanning the target directory for existing ADR files and incrementing the highest found number by one. If no directory is specified or no existing ADRs are found, numbering starts at 0001. </Definition - ADR Numbering>
</Definitions><Goal>
A complete, well-structured ADR file written to the user's chosen location, opened in the session tab for review, with correct sequential numbering and all sections populated from user-provided or user-confirmed content.
</Goal><Rules>
1. Never fabricate options that were not discussed or provided by the user. If context is sparse, ask the user to supply alternatives considered.
2. Always capture rejected alternatives with brief reasoning for why they were not chosen.
3. Past ADRs are immutable. Never modify the body of an existing ADR. To change a past decision, create a new ADR that supersedes it and update only the status line of the old record.
4. Sequential numbering must be determined by scanning the target directory. Never guess or hardcode a number without checking.
5. Every ADR must include a Status field. Default to "Proposed" unless the user explicitly states the decision is already accepted.
6. Never omit the Consequences section. If the user does not volunteer consequences, prompt them to consider at least one positive and one negative implication.
7. When superseding an ADR, always add a forward link in the old record and a backward link in the new record.
8. File names follow the pattern: NNNN-kebab-case-title.md (e.g., 0003-use-postgresql-for-billing.md).
9. Never mix multiple independent decisions into a single ADR. If the user describes two distinct choices, propose splitting into separate records.
10. Always present the draft ADR to the user for review before writing the final file.
</Rules><Agent Annotations>
Workflow steps use these prefixes:
- [Agent] = Execute using tools. Do not involve the user.
- [Ask user] = Present to user and wait for response.
- [Decide] = Evaluate conditions and branch.
- [Think] = Reason internally. Generate candidates, evaluate, select best.
</Agent Annotations><Gotchas>
- Numbering collisions: If the user works across branches, two ADRs may receive the same number independently. After writing, remind the user to verify no collision exists in their main branch before merging.
- Superseded linking: When marking an old ADR as superseded, only the status line and a "Superseded by" link are appended. The rest of the body stays untouched per Rule 3. Read the file, confirm current status, then write back with only the status and link changed.
- Context length: Meeting transcripts can be very long. Summarize the relevant portions rather than embedding the entire transcript in the Context section. If the transcript exceeds 2000 words, extract only the segments pertaining to the decision.
- Template drift: Some teams use modified MADR or Nygard templates with extra sections (e.g., "Compliance Notes"). If the user specifies custom, ask them to describe or paste their template before proceeding.
- Directory assumptions: Never assume the ADR directory is named "docs/adr" or "docs/decisions". Always ask the user where ADRs live, or scan for common patterns in their workspace.
</Gotchas><Instructions><Workflow - Generate ADR description="End-to-end ADR generation from input to written file." tools=[file_write, file_read, run_python, open_in_session_tab] triggers=["create an ADR", "document this architecture decision", "record why we chose X", "write a decision record", "ADR for this design choice"]
- [Ask user] Gather the decision title if not already provided. Confirm the template choice (madr, nygard, or custom). If context was supplied (transcript, notes, or description), acknowledge it. If no context was given, ask the user to describe the decision background in a few sentences. If fails: If the decision title or template choice is missing, re-ask the user for the missing item before continuing.
- [Decide] If the user selected "custom" template, ask them to describe or paste their preferred format. Store the structure for use in the drafting step. Otherwise, proceed with the selected standard template.
- [Think] Analyze the provided context. Extract: the core problem or question, factors that drove the decision (performance, cost, team expertise, timeline), distinct options that were discussed, the chosen option, and any stated trade-offs or risks. If the context is a long transcript, identify only the segments relevant to this specific decision.
[Ask user] Present the extracted elements in a structured summary:
- Problem Statement: (one or two sentences)
- Decision Drivers: (bulleted list)
- Options Considered: (numbered list with one-line descriptions)
- Chosen Option: (name)
- Key Rationale: (why this over the others)
Ask the user to confirm, correct, or add missing items. Per Rule 1, if options are unclear, explicitly ask "Were there other alternatives discussed?" If fails: If the user does not confirm or correct the elements, re-present the summary and ask specifically which items to add or fix.
- [Ask user] Ask about consequences. Per Rule 6, prompt the user: "What positive outcomes do you expect from this decision? What are the downsides or risks?" If the user provides at least one of each, proceed. If not, suggest plausible consequences based on the context and ask for confirmation. If fails: If no consequences are provided or confirmed, propose at least one positive and one negative consequence from the context and ask the user to confirm.
- [Ask user] Confirm the status for this ADR. Default is "Proposed." Ask: "Is this decision already accepted by the team, or should it remain as Proposed for review?" If fails: If the user does not specify a status, default to Proposed and note that the status can be updated later.
- [Ask user] Ask where the ADR file should be saved. Offer to scan the workspace for existing ADR directories. If the user names a path, use it directly. If fails: If no save location is given, offer to scan the workspace for ADR directories and ask the user to pick one.
- [Agent] Scan the target directory for existing ADR files matching the pattern NNNN-*.md. Determine the next sequential number. If the directory does not exist, create it and start at 0001. If fails: If the directory cannot be scanned or created, report the path error to the user and ask for a valid location.
- [Decide] Does this ADR supersede an existing one? If the user indicated it replaces a prior decision, identify the old ADR file. Read its current status. Prepare to update the old file's status to "Superseded" with a forward link per Rule 7.
- [Think] Compose the full ADR document using the selected template. Populate every section with user-confirmed content. Use the appropriate template from the Templates block below. Ensure the file name follows Rule 8.
- [Ask user] Present the complete draft ADR in a code block. Ask: "Does this look correct? Any changes before I write the file?" Wait for explicit approval. Validate: User approves or requests edits. If edits requested, incorporate and re-present. If fails: If the user does not approve or request edits, re-present the draft and ask for explicit approval before writing.
- [Agent] Write the ADR file to the target directory with the computed file name. If this ADR supersedes another, update the old file's status line and append the "Superseded by" link. If fails: If the ADR file cannot be written, report the write error, retry once, and present the draft inline if the retry fails.
- [Agent] Open the newly created ADR in the session tab for the user to review in rendered markdown. If fails: If the ADR cannot be opened in the session tab, report the error and provide the file path so the user can open it manually.
- [Agent] Present a summary: file path, ADR number, status, and a reminder to check for numbering collisions before merging (per Gotchas). If fails: If the summary cannot be presented, report the error and restate the file path and ADR number directly.
</Workflow - Generate ADR>
</Instructions><Templates><Template - MADR>
# {{number}}. {{decision_title}}
Date: {{date}}
## Status
{{status}}
## Context and Problem Statement
{{context_and_problem}}
## Decision Drivers
{{decision_drivers_bulleted}}
## Considered Options
{{options_numbered}}
## Decision Outcome
Chosen option: "{{chosen_option}}", because {{rationale}}.
### Positive Consequences
{{positive_consequences_bulleted}}
### Negative Consequences
{{negative_consequences_bulleted}}
## Links
{{links_to_related_adrs}}</Template - MADR>
<Template - Nygard>
# {{number}}. {{decision_title}}
Date: {{date}}
## Status
{{status}}
## Context
{{context_paragraph}}
## Decision
{{decision_paragraph}}
## Consequences
{{consequences_paragraph}}</Template - Nygard>
</Templates>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.

