WayToClawEarn
Medium impactHacker News

adamsreview: Claude Code multi-lens code review plug-in - a six-layer quality inspection pipeline for the open source community

Open source project adamsreview delivers a six-tier, multi-lens code review pipeline for Claude Code—from parallel sub-agent detection to automated repair loops—that is redefining quality standards for AI code reviews.

WayToClawEarn EditorialPublished May 11, 2026Updated Aug 8, 2026

Editorial review of public sources · AI-assisted drafting. How we work · Original source

Core conclusion

adamsreview is an open source Claude Code plug-in that provides a multi-lens code review pipeline built with six commands. It significantly outperforms Claude Code's native /review, CodeRabbit, and Codex built-in reviews in real PR reviews through parallel sub-agent deployment, automated fix loops, and regression rollback mechanisms.

Key Points

  • Release time: 2026-05-11 (top of Hacker News homepage)
  • Developer: adamjgmiller
  • Technology stack: Shell + Claude Code plug-in system
  • Core breakthrough: parallel multi-camera review + automatic repair + regression rollback

Background

Claude Code, as a representative tool for AI coding agents, provides the basic /review command for PR review. However, there are blind spots in a single-model review perspective—security vulnerability reviews require a different mindset than code correctness reviews. Adamsreview was born for this purpose: it introduces the "multi-person review" concept of the manual team into AI, deploys multiple sub-Agents in parallel (focusing on correctness, security, UX and other dimensions respectively), then merges deduplication, hierarchical verification, and finally unified output.

Core architecture: six-layer pipeline

The six commands of adamsreview constitute a complete code review workflow:

CommandFunctionDescription
/adamsreview:reviewMulti-lens reviewUp to 7 parallel sub-agents, covering dimensions such as correctness/security/UX
/adamsreview:codex-reviewCodex CLI peer reviewSame output format as review, interchangeable with downstream processes
/adamsreview:addExternal discovery injectionInject external results such as Claude Cloud UltraReview, automatically remove duplicates
/adamsreview:walkthroughInteractive item-by-item confirmationDisplay items judged by human eyes one by one, support batch acceptance
/adamsreview:fixAutomatic repair cycleParallel shard repair → Opus secondary review → Regression rollback → Unified submission
/adamsreview:promoteManual upgradeBypass the threshold to force a discovery to be marked as auto-repairable

Key Impact

DimensionsChangeWhat it means to usRecommended actions
Review qualityMore real bugs than native /review and CodeRabbitCI pipeline can significantly reduce manual review workloadTry to introduce multi-lens review in AI-driven projects
WorkflowUpgrade from a single review to a six-layer pipelineCode quality assurance shifts from "manual judgment" to "AI pipeline"Integrate assessment and adamsreview into development workflow
AutomationAutomatic repair + regression rollbackReduce the frequency of manual intervention in repair decisionsEnable automatic repair for non-critical paths
CostBased on Claude Code Max package, no additional deductionsMore economical than /ultrareview (extra deductions)Prioritize using adamsreview instead of UltraReview

Adaptation suggestions

For development teams and AI coding agency users using Claude Code:

  1. Quick experience: Execute /plugin marketplace add adamjgmiller/adamsreview and /plugin install adamsreview@adamsreview in Claude Code
  2. Recommended workflow: For non-trivial PR, execute in the order of review → fix → walkthrough
  3. Ensemble mode: Use --ensemble to enable Claude + Codex CLI dual-engine review at the same time
  4. CI integration: Integrate adamsreview’s automatic repair cycle into the continuous integration process

— AI

Further reading

Tool entry

Tool names appearing in the text: Claude Code, Claude, Codex, OpenAI

Internal link guidance

View source →

Disclaimer: this site shares educational insights only, for inspiration and reference. No outcome guarantee; external execution and decisions are your own responsibility.