Architecture Decision Records
SkillDatabases & dataLets your agent record and look up architecture decisions with reasoning in docs/adr/.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Architecture Decision Records skill
About this capability
Record an architecture decision as an ADR in docs/adr/. Use when choosing between frameworks, libraries, databases, or architectural patterns; stating a decision with reasoning ("we decided X instead of Y because..."); or querying past decisions ("why did we choose X?").
What this skill tells your AI
The instructions your AI receives, as published by linuxfoundation/crowd.dev in .claude/skills/adr/SKILL.md and read by ahel’s review.
You are recording or retrieving an Architecture Decision Record (ADR) for this
project. ADRs live in docs/adr/ at the repo root.
When to record
Record these decisions:
- Technology selections (frameworks, libraries, databases, cloud providers)
- Architectural patterns (state management, caching strategy, API design)
- Data modeling choices (schema design, indexing, query approach)
- Infrastructure and deployment models
- Security, authentication, or testing strategy changes
- Patterns in transition: any documented migration (e.g. Sequelize → pg-promise, classes → functions, multi-tenant → single tenant) is a textbook ADR candidate
Skip: trivial choices (variable naming, formatting, minor refactors).
ADR template
Every ADR file must include all of these sections:
# ADR-NNNN: [Decision Title]
**Date**: YYYY-MM-DD
**Status**: proposed | accepted | deprecated | superseded by ADR-NNNN
**Deciders**: [who was involved]
## Context
[2–5 sentences describing the situation, constraints, and forces at play]
## Decision
[1–3 sentences stating the change clearly and unambiguously]
## Alternatives Considered
### Alternative 1: [Name]
- **Pros**: [benefits]
- **Cons**: [drawbacks]
- **Why not**: [specific rejection reason]
## Consequences
### Positive
- [benefit 1]
### Negative
- [trade-off 1]
### Risks
- [risk and mitigation]
Workflow — recording a new ADR
- Scan existing ADRs —
Glob docs/adr/[0-9]*.mdto find the highest existing number. - Assign next ID — next sequential 4-digit number (e.g.,
0003). - Gather context — ask the user for any missing details: who the deciders were, what alternatives were seriously considered, and what the consequences are.
- Draft the ADR — populate all mandatory sections from the template above.
- Present the draft — show it to the user for review before writing any file.
- Write the file —
docs/adr/NNNN-kebab-title.md(kebab-case title, all lowercase). - Update the index — append a new row to the
| ADR | Title | Status | Date |table indocs/adr/README.md.
Workflow — reading / querying ADRs
- Check if
docs/adr/README.mdexists. If not, offer to start the ADR directory. - Read the README index table and find entries relevant to the user's question.
- Read the matching ADR file and summarise the Context and Decision sections.
- If no ADR matches, suggest recording one now.
Quality standards
- Each ADR should be readable in under 2 minutes.
- Every rejected alternative must include a Why not reason.
- When a decision is superseded, update the old ADR's Status field to
superseded by ADR-NNNNand create the new ADR with a back-reference in its Context. - Keep one decision per ADR.
Signals
- GitHub stars
- 3k
- Forks
- 726
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
adr-linuxfoundation- Source
- github.com/linuxfoundation/crowd.dev