AgentStack
SKILL verified MIT Self-run

Adr

skill-omexit-claude-skills-pack-adr · by OmexIT

Write an Architecture Decision Record (ADR) capturing context, decision, alternatives, and consequences. Triggers: "ADR", "decision record", "why did we choose", "architecture decision".

No reviews yet
0 installs
19 views
0.0% view→install

Install

$ agentstack add skill-omexit-claude-skills-pack-adr

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Adr? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

ADR (Architecture Decision Record)

What I'll do

Capture an architectural decision so future readers understand why the choice was made, what alternatives existed, and what trade-offs were accepted.

When to write an ADR

  • Choosing between frameworks, libraries, or infrastructure
  • Changing a data model, API contract, or system boundary
  • Adopting or abandoning a pattern (monolith → microservices, REST → GraphQL)
  • Any decision that would be hard to reverse and that a new team member would ask "why?"

How I'll think about this

  1. Context over conclusion: Spend more time explaining the forces at play (constraints, requirements, team capabilities, timeline) than the decision itself.
  2. Evaluate alternatives honestly: Each alternative should have real pros — not straw-man options designed to make the chosen one look good.
  3. Name the trade-offs: Every decision has costs. State them explicitly so future readers can re-evaluate if circumstances change.
  4. Keep it short: An ADR is a decision record, not a design doc. If it exceeds 2 pages, consider splitting out the detailed analysis.

Anti-patterns to flag

  • Writing an ADR after implementation just to check a box (context is already lost)
  • Only one alternative listed (signals the decision wasn't actually evaluated)
  • Vague consequences ("this might cause issues later")
  • Missing links to related ADRs or design docs

Quality bar

  • A new team member can read this in 5 minutes and understand why the decision was made
  • Alternatives have genuine pros, not just cons
  • Consequences are concrete and specific
  • Status field is set correctly (Proposed / Accepted / Superseded)
  • Links to related ADRs or design docs if they exist

Workflow context

  • Typically follows: /design-doc (for decisions surfaced during design)
  • Feeds into: /design-doc (as referenced context), /ticket-breakdown
  • Related: /opportunity-assessment (strategic decisions)

Output

Create an ADR using templates/adr.md. Keep it short and specific.

Learning & Memory

After ADR creation completes, save:

  • Decision patterns that recur across the codebase (technology choices, architectural styles, integration approaches)
  • Alternatives considered and why they were rejected, for reuse in future similar decisions
  • Context that informed the decision (constraints, team capabilities, timelines) to detect when circumstances change enough to revisit

Output contract

produces:
  - type: "adr"
    format: "markdown"
    path: "claudedocs/-adr.md"
    sections: [context, decision, alternatives, consequences]
    handoff: "Write claudedocs/handoff-adr-.yaml — suggest: design-doc"

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.