Add new isolated subagent for processing CodeRabbit AI review comments with interactive workflow, severity classification, and verified PR commenting.

[ether-moon]

Features

• Interactive severity-based workflow (CRITICAL/MAJOR/MINOR)
• Pagination support for 200+ GitHub comments
• Language-aware user communication with detection logic
• Triple verification system for PR comment posting
• Context isolation to prevent main session bloat

Implementation

• New agents/ directory architecture
• Agent registration in plugin.json
• Slash command: /skill-set:coderabbit-feedback
• Comprehensive workflow documentation in AGENTS.md

Summary by CodeRabbit

• New Features

  • Added CodeRabbit feedback processor for managing AI review comments with severity-based classification and interactive change application workflow, including PR verification steps.

• Documentation

  • Enhanced documentation with language handling guidance, multilingual support patterns, and Anthropic best practices for development workflows.

[coderabbitai]

Caution

Review failed

The pull request is closed.

Walkthrough

The changes introduce a new CodeRabbit feedback agent workflow with supporting documentation, extend the plugin configuration to register the agent, and enhance AGENTS.md with language handling guidance and Anthropic best practices. The feedback agent processes AI review comments through multi-phase classification, discussion, and application steps.

Changes

Cohort / File(s)	Summary
Documentation & Guidance AGENTS.md	Renamed "Current Skills" to "Current Tools"; added coderabbit-feedback agent entry; reorganized project structure references; introduced "Language Handling in Skills" section with multilingual runtime detection policy; expanded resources section with quality checklists; added Anthropic best practices guidance.
Plugin Configuration skill-set/.claude-plugin/plugin.json	Added new top-level agents field containing entry for ./agents/coderabbit-feedback.md.
New Agent Workflow skill-set/agents/coderabbit-feedback.md	New file defining multi-phase CodeRabbit feedback workflow: Phase 1 (collection and classification of unresolved comments by severity), Phase 2 (interactive discussion and user-driven item selection), Phase 3 (application, verification, and PR comment posting with resolve tag).
New Command skill-set/commands/coderabbit-feedback.md	New command file with front matter and instruction to invoke the coderabbit-feedback agent for processing current PR feedback.

Sequence Diagram

sequenceDiagram
    actor User
    participant Agent as CodeRabbit Feedback<br/>Agent
    participant PR as PR / Git
    participant API as CodeRabbit API

    User->>Agent: Trigger coderabbit-feedback command
    Agent->>API: Fetch unresolved CodeRabbit comments
    API-->>Agent: Return comments with metadata
    Agent->>Agent: Classify by severity<br/>(CRITICAL/MAJOR/MINOR)
    Agent->>User: Present summary & options
    User->>Agent: Select changes to apply<br/>(all or individual review)
    Agent->>Agent: Apply selected changes
    Agent->>PR: Commit & push changes
    Agent->>PR: Post PR comment with<br/>@coderabbitai resolve tag
    Agent->>Agent: Verify comment posted &<br/>contains resolve tag
    Agent-->>User: Workflow complete

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

• AGENTS.md: Extensive documentation updates across multiple sections (language handling, best practices, restructuring) require careful review for accuracy and clarity, though changes are primarily additive.
• coderabbit-feedback.md (agent): Complex multi-phase workflow with detailed state management, verification steps, and user interaction patterns requires validation of logic flow and completeness.
• plugin.json & command file: Configuration and command registration are straightforward but should be verified for correct path references and integration.

Poem

🐰 A fluffy feedback whisker-izer hops in sight,
With phases tri-folded to make PR comments bright,
From CRITICAL to MINOR, each classified with care,
CodeRabbit reviews now get a helpful affair,
Resolve tags and verities—no stone left bare! 🌟

✨ Finishing touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch coderabbit-review-skill

---

📜 Recent review details

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 5e939d1 and 6875dc1.

📒 Files selected for processing (4)

• AGENTS.md (2 hunks)
• skill-set/.claude-plugin/plugin.json (1 hunks)
• skill-set/agents/coderabbit-feedback.md (1 hunks)
• skill-set/commands/coderabbit-feedback.md (1 hunks)

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}