agentleFS
ai-enhanced-engineerai-enhanced-engineer/aiee-skills/skills/arch-decision-records/SKILL.md

arch-decision-records

Skill4 starsChanged 5 months agoMITView on GitHub

Architecture Decision Records (ADRs) for documenting significant technical decisions. Use for capturing decision context, rationale, consequences, and maintaining decision history.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    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
  1. Distinctive:

    One decision per ADR: - Keep focused

    line 44
  2. Distinctive:

    Write when deciding: - Not after the fact

    line 48

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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 Records

Lightweight documentation for architecturally significant decisions.

Core Principle

"Document the why, not just the what."

Code shows what was built; ADRs explain why it was built that way.

What Makes a Decision "Architecturally Significant"?

CriterionExample
Affects system structure"Use event-driven architecture"
Hard to change later"PostgreSQL as primary database"
Impacts non-functional requirements"JWT for authentication"
Involves technology selection"FastAPI over Flask"
Team disagrees or multiple valid optionsAny contested decision

ADR Format

Required Sections

SectionContent
TitleADR-NNN: [Short descriptive title]
StatusProposed / Accepted / Deprecated / Superseded
ContextWhy this decision is needed, constraints, forces
DecisionWhat was decided (clear, unambiguous)
ConsequencesPositive, negative, and neutral outcomes
ReferencesLinks to research, docs, related ADRs

ADR Best Practices

  1. One decision per ADR - Keep focused
  2. Mark as superseded, do not delete - deleting an ADR breaks every inbound reference (PRs, READMEs, other ADRs) and erases the historical reasoning future engineers need to understand the present state
  3. Number sequentially - reusing a number invalidates every existing reference to the prior decision and produces silent collisions in tooling that keys on ADR ID
  4. Keep them short - 1-2 pages maximum
  5. Write when deciding - Not after the fact
  6. Link to research - Traceability to evidence

When NOT to Write an ADR

  • Trivial decisions (naming conventions, formatting)
  • Decisions already covered by team standards
  • Temporary/experimental decisions
  • Decisions with obvious single answers

Status Lifecycle

Proposed → Accepted → [Deprecated | Superseded by ADR-XXX]

RCA & Alternatives Capture

Two ADR-adjacent practices keep decision records auditable: the RCA Verification Matrix (V1/V2 evidence-graded claim list with file/line targets — pause and draft this before prose) and the Alternatives Capture Table (≥2 rejected options with pros/cons/why-not, even when the choice feels obvious). See reference.md → "RCA Verification Matrix" and reference.md → "Capturing Alternatives" for the full patterns and the table schema.

API Consumption Patterns

Prefer the backend's atomic nested-create endpoint (parent + children in one transactional request) over create-parent-then-loop-children. The split form orphans the parent on child failure and duplicates it on retry. Verify the contract first (grep route handlers or curl the schema) before assuming separate endpoints are required.

Anti-PatternPattern
create-parent → loop child createsSingle nested-create; one transaction, no orphan risk
"skip if already created" id-guard on retryMake the create atomic; guard treats symptom, not disease
Assuming split endpoints without checking schemaGrep/curl request schema; nested field may already exist

Server-side rationale lives in arch-ddd § "Aggregate Write Boundaries — Client Side" — cross-reference rather than duplicating transaction semantics here. See reference.md § "API Consumption Patterns" for verification steps.

See reference.md for: Weighted Dimensions / MCDM axis-grounding, Adapter-Lie Smell, Architecture Comparisons Age Unevenly (fast/slow churn taxonomy, CNCF probe), Migration Patterns (migration-before-deletion), Feasibility POC Discipline (four practices).

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.