{
  "source_file": "0930",
  "source_platform": "Dev.to",
  "author": "Mariano Gobea Alcoba",
  "title_listed": "Show HN: adamsreview – better multi-agent PR reviews for Claude Code!",
  "keywords": "",
  "domain": "大模型算法",
  "content_type": "深度技术解析",
  "is_paid_flag": "否",
  "url": "https://dev.to/mgobea/show-hn-adamsreview-better-multi-agent-pr-reviews-for-claude-code-3pb3",
  "publish_date": "2026-05-11",
  "corpus_stage": "",
  "notes": "",
  "id": "5ca2eb71f591",
  "collected_at": "2026-10-02T14:05:40",
  "title": "Show HN: adamsreview – better multi-agent PR reviews for Claude Code! - DEV Community",
  "body_markdown": "Advanced Multi-Agent System for Enhanced Code Review with Claude Code\nThe proliferation of AI-assisted code review tools has introduced novel paradigms for identifying defects and improving code quality. While existing solutions like Claude Code's built-in\n/review\nand\n/ultrareview\ncommands, alongside third-party offerings such as CodeRabbit and Greptile, provide valuable automation, they often operate under a single-pass, monolithic review model. This approach can limit their ability to perform in-depth analysis, manage complex dependencies, and effectively integrate human feedback. This article details the design and implementation of\nadamsreview\n, a Claude Code plugin engineered to address these limitations by leveraging a multi-agent, multi-stage review process.\nadamsreview\nis conceived as a system of interconnected sub-agents, orchestrated to perform distinct analytical tasks. This architecture allows for a more granular and robust review process, moving beyond the capabilities of simpler, single-pass AI reviews. The core philosophy is to decompose the review into manageable stages, each handled by specialized agents, with explicit state management and mechanisms for human intervention and iterative refinement.\nSystem Architecture and Core Components\nThe\nadamsreview\nplugin comprises six distinct Claude Code slash commands, each representing a stage or utility within the review workflow:\n/review\n: Initiates a comprehensive, multi-stage review process.\n/codex-review\n: Integrates with Codex CLI for an ensemble review approach, augmenting Claude's analysis.\n/add\n: Allows for the explicit inclusion of specific files or directories in the review scope.\n/promote\n: Facilitates the promotion of specific findings to higher stages of review or action.\n/walkthrough\n: Engages Claude's\nAskUserQuestion\nfeature to present uncertain findings or items requiring human judgment iteratively.\n/fix\n: Orchestrates the resolution of identified issues, including group-based agent dispatch and regression testing.\nA key architectural tenet is the management of review state. Unlike ephemeral review processes,\nadamsreview\nutilizes persistent JSON artifacts stored on disk. This state management is crucial for enabling multi-stage reviews where context can be cleared between stages without losing critical information. Scripts are included to manage the lifecycle of this state, ensuring data integrity and facilitating subsequent review iterations.\nMulti-Stage Review Process\nThe primary\n/review\ncommand is the entry point to the multi-stage process. It initiates a series of parallel sub-agent analyses, followed by a sequential validation pass.\nParallel Sub-Agent Analysis\nUpon invocation,\n/review\ntriggers an array of specialized Claude Code agents to operate in parallel. These agents are tasked with specific aspects of code analysis:\nSecurity Agent\n: Scans for common security vulnerabilities (e.g., SQL injection, XSS, improper authentication).\nPerformance Agent\n: Identifies potential performance bottlenecks (e.g., inefficient loops, redundant computations, suboptimal data structures).\nMaintainability Agent\n: Assesses code readability, complexity, and adherence to design principles (e.g., SOLID, DRY).\nBug Detection Agent\n: Focuses on identifying logical errors, off-by-one errors, null pointer dereferences, and other common programming mistakes.\nStyle Agent\n: Enforces coding style guidelines and best practices.\nEach of these agents operates independently, processing the provided code context. The results are aggregated, and a preliminary report is generated.\nSequential Validation Pass\nFollowing the parallel analysis, a sequential validation pass is performed. This stage involves a more holistic evaluation of the aggregated findings. A dedicated \"Validator Agent\" reviews the output from the parallel sub-agents, looking for:\nFalse Positives\n: Cross-referencing findings to identify redundant or incorrect reports.\nInterdependencies\n: Analyzing how findings in one area might impact another.\nSeverity Prioritization\n: Assigning severity levels (e.g., Critical, High, Medium, Low) to identified issues based on potential impact.\nThis validation pass aims to refine the raw output from the sub-agents, producing a more coherent and actionable review report.\nState Management and Context Persistence\nThe persistence of review state through JSON artifacts is a distinguishing feature of\nadamsreview\n. This mechanism allows for:\nIntermediate State Saving\n: After each significant stage of the review, the state is serialized to a JSON file. This file typically includes the code diff, the aggregated findings from previous stages, and any user-provided annotations.\nContextual Clarity Between Stages\n: When a user invokes a subsequent command (e.g.,\n/walkthrough\nafter\n/review\n), the system loads the relevant JSON state. This ensures that the AI has access to the historical findings and the current state of the review, even if the intermediate Claude Code session context has been cleared.\nSelective Review Scope\n: The\n/add\ncommand allows users to augment the review scope with specific files or directories. This information is appended to the persistent state, ensuring that future review stages consider the expanded scope.\nState Management Scripts\n: Utility scripts are provided to manage the creation, updating, and clearing of these JSON state files, offering a programmatic interface for controlling the review lifecycle.\nThe JSON state might adopt a structure similar to this:\n{\n\"commit_hash\"\n:\n\"a1b2c3d4e5f67890\"\n,\n\"base_branch\"\n:\n\"main\"\n,\n\"review_files\"\n:\n[\n\"src/utils.py\"\n,\n\"src/models.py\"\n],\n\"findings\"\n:\n[\n{\n\"stage\"\n:\n\"initial_analysis\"\n,\n\"agent\"\n:\n\"security_agent\"\n,\n\"file\"\n:\n\"src/models.py\"\n,\n\"line\"\n:\n42\n,\n\"message\"\n:\n\"Potential SQL injection vulnerability in user_query function.\"\n,\n\"severity\"\n:\n\"High\"\n,\n\"details\"\n:\n\"The user input is directly concatenated into the SQL query string without sanitization.\"\n},\n{\n\"stage\"\n:\n\"initial_analysis\"\n,\n\"agent\"\n:\n\"performance_agent\"\n,\n\"file\"\n:\n\"src/utils.py\"\n,\n\"line\"\n:\n105\n,\n\"message\"\n:\n\"Inefficient loop detected in data_processing function.\"\n,\n\"severity\"\n:\n\"Medium\"\n,\n\"details\"\n:\n\"Consider using a vectorized operation instead of iterating through each element.\"\n}\n],\n\"user_annotations\"\n:\n[],\n\"review_status\"\n:\n\"in_progress\"\n}\nEnter fullscreen mode\nExit fullscreen mode\nHuman-AI Collaboration and Iterative Refinement\nadamsreview\nplaces a strong emphasis on facilitating human-AI collaboration, particularly in handling uncertainty and driving towards resolution.\n/walkthrough\nCommand\nThe\n/walkthrough\ncommand is designed to address findings that are potentially ambiguous or require domain-specific knowledge that the AI might not fully possess. It leverages Claude's\nAskUserQuestion\nfeature to interactively engage the user:\nPresentation of Findings\n: The command iterates through the aggregated findings from the persistent state.\nInteractive Querying\n: For each finding deemed to require human judgment (e.g., based on confidence scores or pre-defined heuristics),\nadamsreview\nuses\nAskUserQuestion\nto present the finding to the user.\nUser Feedback Loop\n: The user can then provide feedback, ask clarifying questions, or instruct the AI on how to proceed. This interaction is recorded and incorporated back into the persistent state.\nIterative Refinement\n: This process can be repeated, allowing users to progressively refine the review results and guide the AI's understanding.\nThis interactive approach transforms the review from a black-box process into a dynamic dialogue.\n/promote\nCommand\nThe\n/promote\ncommand allows users to explicitly elevate the importance of certain findings. This can be useful for:\nMarking Critical Issues\n: Users can mark specific findings as \"critical\" or \"must-fix\" regardless of the AI's initial severity assessment.\nContextualizing Findings\n: Users can add additional context or justifications to findings, which can then be used by subsequent agents or for reporting.\nThe promoted findings are updated in the persistent JSON state, influencing subsequent review or fix stages.\nEnsemble Review with Codex CLI\nThe\n/codex-review\ncommand introduces an ensemble approach by integrating with the Codex CLI. This offers an alternative or complementary review perspective:\nCode Export\n: The relevant code diff or subset of files is exported in a format compatible with Codex CLI.\nCodex CLI Execution\n: The Codex CLI is invoked with specific prompts designed to elicit code review feedback.\nResult Aggregation\n: The output from Codex CLI is parsed and merged with the findings from Claude's native review.\nCross-Validation\n: This ensemble approach enables cross-validation of findings. If both Claude and Codex identify a similar issue, the confidence in that finding increases. Discrepancies can highlight areas where one model might be stronger than the other or where an issue is particularly subtle.\nThis strategy aims to leverage the strengths of different AI models, potentially reducing the false positive rate and increasing the detection of more nuanced bugs.\nAutomated Fixing and Regression Prevention\nThe\n/fix\ncommand is designed to automate the remediation of identified issues, incorporating a robust process for preventing regressions.\nPer-Fix-Group Agent Dispatch\nIssues are often related. For instance, a security vulnerability might necessitate changes across multiple files, or a refactoring effort might span several related functions. The\n/fix\ncommand groups related findings together. For each identified \"fix group\":\nSpecialized Fix Agent\n: A dedicated \"Fix Agent\" is dispatched. This agent is tasked with understanding the scope of the fix group and proposing code modifications.\nIterative Fixing\n: The agent may iterate on its proposed fixes, attempting to resolve all issues within the group.\nCommit Planning\n: Proposed changes are staged for review.\nRe-Review and Regression Testing\nAfter the Fix Agent has proposed modifications,\nadamsreview\nperforms a crucial re-review and regression check:\nPost-Fix Review\n: The modified code is immediately subjected to a subset of the original review agents (particularly the bug detection and security agents). This \"post-fix review\" aims to identify any new issues introduced by the attempted fixes (regressions).\nUnit Test Execution (Optional but Recommended)\n: If a testing framework is integrated with the development environment,\nadamsreview\ncan trigger unit tests. This provides a more direct measure of functional correctness.\nSurvivor Commit\n: Only changes that pass the post-fix review and all executed tests are committed. Findings that introduce regressions or new issues are reverted.\nIterative Fix Attempt\n: If fixes are reverted, the findings associated with those fixes are returned to the persistent state, potentially with updated information from the regression analysis, allowing for further attempts at remediation.\nThis disciplined approach ensures that automated fixes are safe and do not compromise existing code quality.\nComparison with Existing Tools\nadamsreview\ndistinguishes itself from existing solutions in several key aspects:\n/review\nvs.\n/ultrareview\n: While\n/ultrareview\nin Claude Code offers enhanced capabilities, it draws from the \"Extra Usage\" pool, incurring direct costs.\nadamsreview\noperates on a standard Claude Code subscription (Max plan recommended for extensive context windows), providing a more cost-effective, deeper review.\nDepth of Analysis\n: By employing a multi-stage, multi-agent approach with parallel sub-analyses and explicit validation,\nadamsreview\naims for a more comprehensive detection rate of bugs and vulnerabilities compared to single-pass tools.\nState Persistence\n: The explicit JSON state management enables multi-stage reviews and context continuity, which is not a standard feature in many AI review tools that often operate within a single conversational turn or ephemeral session.\nHuman-AI Collaboration\n: The\n/walkthrough\ncommand, using\nAskUserQuestion\n, provides a structured way for humans to guide and validate AI findings, fostering a more collaborative development process.\nEnsemble Capabilities\n: The\n/codex-review\ncommand's integration with Codex CLI offers an ensemble review perspective, potentially improving accuracy and reducing false positives.\nAutomated Fix and Regression Prevention\n: The\n/fix\ncommand's structured approach to fixing issues, including post-fix re-reviews and regression checks, provides a more robust automated remediation process than simple patch generation.\nImplementation Details and Usage\nThe\nadamsreview\nplugin is installed using Claude Code's plugin marketplace:\n/plugin marketplace add adamjgmiller/adamsreview\n/plugin\ninstall\nadamsreview@adamsreview\nEnter fullscreen mode\nExit fullscreen mode\nExample Workflow:\nInitiate Review\n:\n/review\nThis triggers the multi-stage analysis. Findings are stored in a JSON artifact.\nAdd Specific Files (Optional)\n: If the initial review missed certain critical files, or if the user wants to ensure specific files are considered in subsequent stages:\n/add src/config/settings.py tests/unit/test_api.py\nThe state is updated to include these files.\nInteractive Walkthrough\n: For findings that require user input:\n/walkthrough\nClaude Code prompts the user with questions about specific findings. User responses update the state.\nPromote a Finding\n: If a user identifies a finding as particularly critical:\n/promote finding_id_123\n--priority\ncritical\n--comment\n\"This is a major security flaw.\"\nThe finding's metadata is updated in the state.\nEnsemble Review (Optional)\n: To augment Claude's analysis with Codex:\n/codex-review\nCodex CLI is invoked, and its findings are merged into the state.\nAutomated Fix Attempt\n: To fix identified issues:\n/fix\nAgents attempt to fix issues, followed by a re-review and regression check. Commits are made only for safe fixes.\nClearing State\n: To start a fresh review, the JSON state file needs to be removed or managed by the utility scripts.\nThe recommended plan for using\nadamsreview\neffectively is Claude Code's Max plan, which typically offers larger context windows. This is beneficial for processing extensive codebases and detailed diffs, which are common in complex PRs, thereby maximizing the effectiveness of the multi-agent system.\nFuture Enhancements and Considerations\nCustomizable Agent Configurations\n: Allowing users to enable/disable specific sub-agents or tune their parameters.\nIntegration with CI/CD Pipelines\n: Enabling\nadamsreview\nto be triggered automatically as part of a CI/CD workflow.\nAdvanced Regression Detection\n: Incorporating more sophisticated static analysis tools or fuzzing techniques for regression detection.\nLearning from User Feedback\n: Developing mechanisms for the AI to learn from user annotations and correction patterns over time.\nBroader LLM Integration\n: Extending the ensemble review to include other large language models.\nConclusion\nadamsreview\npresents a robust and extensible framework for AI-assisted code review, designed to overcome the limitations of simpler, monolithic approaches. By employing a multi-stage, multi-agent architecture with sophisticated state management, human-AI collaboration features, and automated regression prevention, it aims to deliver significantly more accurate and actionable insights than existing tools. The system's modular design allows for continuous improvement and adaptation, paving the way for more intelligent and collaborative code review processes.\nFor organizations seeking to enhance their code quality and streamline their development workflows through advanced AI-driven code review solutions, consulting services can be invaluable. Visit\nhttps://www.mgatc.com\nto explore how expert guidance can help implement and optimize such sophisticated systems within your development lifecycle.\nOriginally published in Spanish at\nwww.mgatc.com/blog/adamsreview-better-multi-agent-pr-reviews-for-claude-code/",
  "metadata": {
    "title": "Show HN: adamsreview – better multi-agent PR reviews for Claude Code! - DEV Community",
    "description": "Advanced Multi-Agent System for Enhanced Code Review with Claude Code   The proliferation of... Tagged with claudecode, ai, codereview, softwaredevelopment.",
    "keywords": "claudecode, ai, codereview, softwaredevelopment, software, coding, development, engineering, inclusive, community",
    "author": null,
    "og:type": "article",
    "og:url": "https://dev.to/mgobea/show-hn-adamsreview-better-multi-agent-pr-reviews-for-claude-code-3pb3",
    "og:title": "Show HN: adamsreview – better multi-agent PR reviews for Claude Code!",
    "og:description": "Advanced Multi-Agent System for Enhanced Code Review with Claude Code   The proliferation of...",
    "og:site_name": "DEV Community",
    "og:image": "https://media2.dev.to/dynamic/image/width=1200,height=627,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fplv304cluel1jxohs1og.png",
    "twitter:site": "@thepracticaldev",
    "twitter:creator": "@",
    "twitter:title": "Show HN: adamsreview – better multi-agent PR reviews for Claude Code!",
    "twitter:description": "Advanced Multi-Agent System for Enhanced Code Review with Claude Code   The proliferation of...",
    "twitter:card": "summary_large_image",
    "twitter:widgets:new-embed-design": "on",
    "twitter:image:src": "https://media2.dev.to/dynamic/image/width=1200,height=627,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fplv304cluel1jxohs1og.png"
  },
  "paywall": {
    "listed_paid": false,
    "detected": false,
    "marker": "",
    "public_only": false,
    "policy": "付费/会员内容只保留页面公开可见部分，不绕过付费墙"
  },
  "crawl": {
    "success": true,
    "status_code": 200,
    "error": ""
  },
  "code_blocks": [
    {
      "file": "attachments/5ca2eb71f591/code/body-000.txt",
      "chars": 891,
      "language": "text"
    },
    {
      "file": "attachments/5ca2eb71f591/code/body-001.txt",
      "chars": 89,
      "language": "text"
    },
    {
      "file": "attachments/5ca2eb71f591/code/body-002.txt",
      "chars": 8,
      "language": "text"
    },
    {
      "file": "attachments/5ca2eb71f591/code/body-003.txt",
      "chars": 51,
      "language": "text"
    },
    {
      "file": "attachments/5ca2eb71f591/code/body-004.txt",
      "chars": 13,
      "language": "text"
    },
    {
      "file": "attachments/5ca2eb71f591/code/body-005.txt",
      "chars": 87,
      "language": "text"
    },
    {
      "file": "attachments/5ca2eb71f591/code/body-006.txt",
      "chars": 14,
      "language": "text"
    },
    {
      "file": "attachments/5ca2eb71f591/code/body-007.txt",
      "chars": 5,
      "language": "text"
    }
  ],
  "images": [],
  "audio": [],
  "videos": [],
  "links": [
    {
      "href": "https://www.mgatc.com/",
      "label": "https://www.mgatc.com",
      "type": "a"
    },
    {
      "href": "https://www.mgatc.com/blog/adamsreview-better-multi-agent-pr-reviews-for-claude-code/",
      "label": "www.mgatc.com/blog/adamsreview-better-multi-agent-pr-reviews-for-claude-code/",
      "type": "a"
    }
  ],
  "repair_version": 2,
  "repaired_at": "2026-10-14T10:50:04Z",
  "source_url": "https://dev.to/mgobea/show-hn-adamsreview-better-multi-agent-pr-reviews-for-claude-code-3pb3",
  "papers": [],
  "actual_video_files": [],
  "body_chars": 15961,
  "expected_body_images": 0,
  "discovery_status": "ok",
  "delivery_status": "ok",
  "missing_reason": "",
  "raw_html": "attachments/5ca2eb71f591/pages/source.html",
  "local_html": "attachments/5ca2eb71f591/pages/article.html",
  "body_base_path": "attachments/5ca2eb71f591/pages/",
  "body_markdown_path": "attachments/5ca2eb71f591/pages/article.md"
}