adr-editor

Reviews and curates Architecture Decision Records through ranked clusters and maintainer-approved dispositions.

20|2|Updated May 16, 2026
One-click install
npx skills add https://github.com/reddb-io/red-skills --skill adr-editor-reddb-io
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: adr-editor
Source: https://github.com/reddb-io/red-skills/tree/main/plugins/dev/skills/engineering/adr-editor
Command: npx skills add https://github.com/reddb-io/red-skills --skill adr-editor-reddb-io

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Architecture Decision Record collections accumulate decision debt over time: records go stale, contradict current code, duplicate each other, or become obsolete without anyone noticing. This Skill systematically confronts each ADR cluster with current implementation evidence and applies maintainer-approved changes so the decision history stays accurate. ## Core Features & Use Cases - Ranked Cluster Triage: Inventories active ADRs in .red/adr/, groups them into clusters, detects inconsistencies, and recommends where to start based on fresh evidence and review markers. - Proposal-Driven Reverse Grill: Presents one numbered proposal at a time (P01, P02, ...) with evidence, exact operations, alternatives, and a recommendation, requiring an explicit maintainer disposition for every active record. - Full ADR Verb Set: Supports list, group, surface inconsistencies, add, remove, rewrite, merge, split, archive, renumber, and re-index through deterministic planner/apply helpers with rollback. - Use Case: A maintainer suspects several ADRs no longer match the shipped code. The Skill ranks the clusters, confronts one cluster against current code and tests, walks the maintainer through dispositions, previews the exact diff, and lands one coherent cluster change per PR. ## Quick Start Ask the agent to review the active ADR collection, rank the clusters, and recommend which cluster to confront first.

Frequently Asked Questions about adr-editor

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

FAQPage Schema
How do I review and clean up stale Architecture Decision Records?▼

Run the ADR editor workflow: it inventories active records in .red/adr/, ranks clusters by fresh evidence, then confronts one cluster against current code, tests, and docs. Each record gets an explicit maintainer disposition before any change is applied.

How to merge or split ADRs without losing decision history?▼

Use the merge, split, or absorb operations. Merge mints a successor ADR and archives all originals; absorb rewrites one governing ADR and archives only auxiliaries; split creates focused successors. All archived records keep terminal status and successor pointers.

Can the ADR editor modify product source code?▼

No. The workflow edits ADR support artifacts only, such as ADR files and .red/adr/INDEX.md. A mismatch between an ADR and the product is treated as evidence for a proposal, never as authority to repair product code in the same PR.

When should an ADR be archived instead of rewritten?▼

Archive only when current evidence and maintainer judgment show the record is auxiliary, deprecated, superseded, absorbed, or otherwise terminal. Age or missing inbound links alone are never sufficient archival evidence.

Does the ADR editor apply changes automatically without confirmation?▼

No. It accumulates accepted proposals, shows the complete resulting text and exact diff for every affected file, then asks once for confirmation before applying the destructive batch. Stop, silence, or requested edits cancel the plan.