adr-critique

Audit ADRs for filler, hedging, and drift with line-level rewrite proposals.

12|Updated Apr 19, 2026
One-click install
npx skills add https://github.com/janmohammadi/deep-adr --skill adr-critique
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: adr-critique
Source: https://github.com/janmohammadi/deep-adr/tree/main/skills/adr-critique
Command: npx skills add https://github.com/janmohammadi/deep-adr --skill adr-critique

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Audit existing or legacy ADRs that weren't written through /draft-adr — flag filler, hedging, implementation detail, missing why, weak Architecture Contracts, inconsistency with other ADRs, and LikeC4 drift. Produces line-level rewrite suggestions for human review.

Core Features & Use Cases

  • Quote the exact original line and identify the violation, then propose a tighter rewrite for human approval.
  • Detect Missing-why, weak Architecture Contracts, and drift relative to LikeC4 models to ensure consistency and traceability.
  • Facilitate quality governance over ADR corpora by isolating surface issues from substantive decision content.

Quick Start

Run the adr-critique skill on an existing ADR to generate line-by-line rewrite suggestions for human review.

Frequently Asked Questions about adr-critique

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit existing architecture decision records for filler and missing context?

Auditing existing architecture decision records (ADRs) for filler and missing context involves analyzing legacy documents to surface hedging, implementation details, and missing rationale, then generating line-level rewrite proposals for human review and approval.

What is ADR drift and how do I detect inconsistency in architecture decisions?

ADR drift occurs when documented architecture decisions become inconsistent with actual system models or other ADRs. You can detect this drift by auditing ADRs against LikeC4 models to ensure consistency, traceability, and alignment with current architecture contracts.

How do I rewrite legacy ADRs to improve quality and precision?

Rewriting legacy ADRs to improve precision requires isolating surface issues like weak architecture contracts from substantive decision content, then applying line-by-line rewrite suggestions that tighten language and restore missing rationale for human approval.

Can I check imported ADRs for weak architecture contracts and missing why?

Yes, you can check imported ADRs for weak architecture contracts and missing rationale by running an audit that adheres to ADR-IS-NOT rules, quoting exact original lines that violate these rules, and flagging where explicit reasoning is lacking for human review.

What are the limitations of automated ADR critique and when should I review manually?

Automated ADR critique is limited to surfacing surface-level quality issues like filler and drift; it requires manual human review to approve per-line rewrite suggestions, ensuring substantive decision content remains intact and accurate to the actual architecture.