ADR

SkillDev tools

Capture qualifying architecture decisions as ADRs in the right HQ or repo location.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the ADR skill

What this skill tells your AI

The instructions your AI receives, as published by indigoai-us/hq-core in .claude/skills/adr/SKILL.md and read by ahel’s review.

Architectural Decision Records — minimal, dated, sequentially numbered. The point is recording that a decision was made and why, not filling out template sections.

Pattern adapted from mattpocock/skills (grill-with-docs/ADR-FORMAT.md).

Step 0 — Resolve company + scope

$ARGUMENTS may include:

  • [company] slug — fall back to manifest / handoff / cwd inference if absent
  • [repo|hq] scope — pick home directory
  • <one-line decision summary> — used as title and slug seed

If scope is ambiguous, ask via AskUserQuestion:

ChoiceHomeWhen
repo<repo>/docs/adr/Decision is bound to specific code; reviewers will read it alongside the code
hqcompanies/{co}/knowledge/adrs/Decision is org-level (e.g. "we use Linear, not Jira"; "all auth via Cognito")

If no company resolves and scope is hq, default to core/knowledge/public/adrs/ (rare — most ADRs have a tenant).

Step 1 — Three-condition gate (HARD BLOCK)

Ask the user three yes/no questions via a single AskUserQuestion call:

  1. Hard to reverse — would changing your mind later cost meaningful effort? (refactor, migration, contract renegotiation)
  2. Surprising without context — would a future reader look at the result and wonder "why on earth did they do it this way?"
  3. Result of a real trade-off — were there genuine alternatives, and did you pick this one for specific reasons?

If any answer is no → DO NOT WRITE THE ADR. Offer alternatives:

Failed conditionAlternative
Easy to reverseSkip — you'll just reverse it later
Not surprisingSkip — nobody will wonder
No real trade-offAdd to CONTEXT.md glossary instead, or skip

Re-prompt: "Skip ADR / Add to CONTEXT.md / Force-write anyway (rarely correct)". Only Force-write proceeds; record reasoning in the ADR body so the override is explicit.

Step 2 — Locate or create target directory

# repo scope
mkdir -p <repo>/docs/adr/

# hq scope
mkdir -p companies/{co}/knowledge/adrs/

Lazy-create only when the first ADR is needed.

Step 3 — Compute next number

ls <target>/[0-9][0-9][0-9][0-9]-*.md 2>/dev/null \
  | sed 's|.*/\([0-9]*\)-.*|\1|' \
  | sort -n \
  | tail -1

Increment by 1. Default to 0001 if no existing ADRs. Format: NNNN (4-digit zero-padded).

Slug from the user's title:

  • Lowercase
  • Spaces / underscores → -
  • Strip non-alphanumeric except -
  • Cap at 60 chars

Final filename: NNNN-slug.md

Step 4 — Walk the user through 1–3 sentences

Single AskUserQuestion (or sequential prompts) collecting:

  • Title — short noun phrase, sentence case, ≤60 chars
  • Body — 1–3 sentences answering: what's the context, what did we decide, why this over alternatives

That's the minimum. The value is in recording the decision, not in completing a template.

Step 5 — Optional sections (offer, don't force)

Ask via AskUserQuestion (multiSelect) which (if any) to include:

  • Status frontmatter (proposed | accepted | deprecated | superseded by ADR-NNNN) — useful when decisions are likely to be revisited
  • Considered Options — only when rejected alternatives are worth remembering
  • Consequences — only when non-obvious downstream effects need calling out

If user picks none, the ADR is just title + 1–3 sentences. Good.

Step 6 — Write ADR

Template (use only the sections selected in Step 5):

---
status: <accepted|proposed|...>          # only if Status was selected
date: <YYYY-MM-DD>
related: []                              # ADR numbers, optional
---

# <Title>

<1-3 sentences: context, decision, why.>

## Considered Options                    <!-- optional -->

- <Option A> — <why rejected>
- <Option B> — <why rejected>

## Consequences                          <!-- optional -->

- <non-obvious downstream effect>

Step 7 — Cross-link

If new domain terms surfaced during the conversation, offer to update CONTEXT.md (in the same repo for repo scope; in companies/{co}/knowledge/CONTEXT.md for hq scope).

If this ADR supersedes an earlier one, update the older ADR's frontmatter to status: superseded by ADR-NNNN (after asking).

If this ADR is accepting a /architect candidate, add a backlink in workspace/reports/{slug}-architect.md under that candidate's "Outcome" section.

Step 8 — Report

Print:

✓ ADR written: <target>/NNNN-slug.md
  Title: <title>
  Scope: repo | hq
  Optional sections: <Status / Considered Options / Consequences / none>
  CONTEXT.md updated: yes / no
  Supersedes: ADR-MMMM / none

What qualifies (examples)

  • Architectural shape. "We're using a monorepo." "Write model is event-sourced; read model projects to Postgres."
  • Integration patterns between contexts. "Ordering and Billing communicate via domain events, not synchronous HTTP."
  • Technology choices that carry lock-in. Database, message bus, auth provider, deployment target. Not every library — just the ones that would take a quarter to swap out.
  • Boundary and scope decisions. "Customer data is owned by the Customer context; other contexts reference it by ID only." Explicit no-s are as valuable as the yes-s.
  • Deliberate deviations from the obvious path. "We're using manual SQL instead of an ORM because X." Anything where a reasonable reader would assume the opposite. Stops the next engineer from "fixing" something deliberate.
  • Constraints not visible in the code. "We can't use AWS because of compliance requirements." "Response times must be under 200ms because of the partner API contract."
  • Rejected alternatives when the rejection is non-obvious. Considered GraphQL, picked REST for subtle reasons → record it, otherwise someone will suggest GraphQL again in six months.

What does NOT qualify

  • Library choices that are easy to swap (date formatter, test runner version, lint rule preference)
  • Naming decisions ("we call this thing the X intake") → these go in CONTEXT.md, not an ADR
  • Implementation details that the code itself makes obvious
  • Decisions you're already going to revisit next sprint (use a TODO comment)
  • "We did the obvious thing" — there's nothing to record

Cross-references

  • /out-of-scope — sibling skill for rejected feature requests (vs. this for accepted technical decisions).
  • /brainstorm, /plan, /architect, /diagnose — all hand off here when a decision warrants ADR.
  • HQ /learn — for cross-tenant knowledge capture (different from per-repo / per-company ADR).
  • Pattern source: mattpocock/skills grill-with-docs/ADR-FORMAT.md.

Signals

GitHub stars
84
Forks
15
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
adr-indigoai-us
Source
github.com/indigoai-us/hq-core