Skip to content

Latest commit

 

History

History
432 lines (321 loc) · 10.8 KB

File metadata and controls

432 lines (321 loc) · 10.8 KB

Developer Plugin

A comprehensive toolkit for developers including specialized agents for code review, security analysis, and code quality management.

🤖 Available Agents

Code Reviewer

Command: @code-reviewer Color: 🟢 Green

Expert code reviewer for bugs, logic errors, security vulnerabilities, and project guidelines compliance. Uses confidence-based filtering (≥80%) to report only high-priority issues.

Use cases:

  • Review code before pull requests
  • Check adherence to project conventions (CLAUDE.md)
  • Identify bugs and security issues
  • Ensure code quality standards

Security Agent

Command: @code-security Color: 🔴 Red

Pragmatic application security engineer for multi-language code review with automated scanner integration.

Features:

  • Cross-language security analysis (Python, Go, PHP, JavaScript, etc.)
  • Automated security scanning (bandit, gosec, semgrep, npm audit, etc.)
  • Dependency vulnerability checking
  • CWE/CVE identification with actionable fixes

Use cases:

  • Security audits before deployment
  • Check for common vulnerabilities
  • Audit dependencies for CVEs
  • Track security posture over time

Code Mess Detector

Command: @code-mess-detector Color: 🟡 Yellow

Analyzes code written during rapid prototyping ("vibe-coding") for common quality issues. Generates detailed reports for systematic cleanup.

Detects:

  • Inconsistent naming and poor structure
  • Missing error handling and documentation
  • Code duplication and magic numbers
  • Dead code and debug statements
  • TODO/FIXME comments

Output:

  • JSON report: .audit/agents/mess-detector/report.json
  • Markdown summary: .audit/agents/mess-detector/summary.md

Use cases:

  • Clean up after rapid prototyping
  • Identify technical debt
  • Prepare code for production
  • Maintain code quality

Code Mess Fixer

Command: @code-mess-fixer Color: 🔵 Blue

Systematically applies fixes based on code-mess-detector reports. Works through issues by priority and tracks progress.

Features:

  • Reads detection reports and prioritizes by severity
  • Applies safe, targeted fixes automatically
  • Tracks progress and updates report status
  • Skips complex changes requiring manual review

Use cases:

  • Automated cleanup after detection
  • Systematic technical debt reduction
  • Safe refactoring with progress tracking

⚡ Slash Commands

Add Task

Command: /add-task <task description>

Quickly add a task to the task list without interrupting your current work context.

Features:

  • Creates a pending task using TaskCreate
  • Immediately continues with current work
  • Keeps focus on what you're doing while noting things for later

Usage:

# Add a task while working on something else
/add-task Fix the failing unit test in auth module

# Note something to come back to later
/add-task Refactor the database connection handling

Create PR

Command: /create-pr [base-branch]

Analyzes your changes and creates a pull request with AI-generated title and description.

Features:

  • Analyzes commit history and diff since base branch
  • Generates comprehensive PR title and description
  • Automatically pushes branch if needed
  • Creates PR using GitHub CLI (gh)
  • Follows conventional commit format
  • Includes testing notes and breaking changes

Usage:

# Create PR against default branch
/create-pr

# Create PR against specific branch
/create-pr develop

# Create draft PR
/create-pr main draft

Prerequisites:

  • GitHub CLI (gh) installed and authenticated
  • Branch has commits not in base branch

Handle Conflicts

Command: /handle-conflicts

Assists with resolving git merge conflicts during merge, rebase, or cherry-pick operations.

Features:

  • Detects operation type (merge, rebase, cherry-pick)
  • Lists all conflicted files
  • Provides context about current branch and commits being applied
  • Guides through systematic conflict resolution
  • Supports reviewing changes for future auditing

Usage:

# When you have merge conflicts
/handle-conflicts

Prerequisites:

  • Active merge/rebase/cherry-pick with conflicts

Process PR Reviews

Command: /gh-process-review [continue]

Systematically process GitHub PR review comments by fetching threads to local JSON, implementing fixes one-by-one, and tracking progress.

Features:

  • Single CLI entry point (review.sh) with subcommands: fetch, list, get, reply, mark
  • Auto-detects PR and reviews file from current branch — no manual path threading
  • Fetches review threads via GraphQL (includes resolution status)
  • Stores reviews locally in .scratch/reviews/ to avoid context pollution
  • Lists unresolved threads with smart filtering (excludes resolved/outdated)
  • Posts "Fixed in <hash>" replies to threads automatically
  • Tracks local resolution status for progress tracking
  • Supports "continue" mode to resume with existing reviews file

Usage:

# Process reviews for current branch's PR
/gh-process-review

# Resume processing with existing reviews file
/gh-process-review continue

Workflow:

  1. Fetches all review threads to local JSON
  2. Lists unresolved threads needing attention
  3. For each thread: read feedback → implement fix → commit → push → reply with commit hash
  4. Marks threads as locally resolved for tracking

Prerequisites:

  • GitHub CLI (gh) installed and authenticated
  • Current branch has an associated PR with review comments
  • jq installed for JSON processing

🔌 MCP Servers

Linear

Integration with Linear for issue tracking and project management.

Features:

  • Create, read, update Linear issues
  • Search and filter issues
  • Manage project workflows
  • Track issue status and assignments

Setup:

  1. Plugin automatically configures the Linear MCP
  2. Run /mcp to authenticate with Linear (OAuth)
  3. Start using Linear tools in your workflow

📊 Context Window Progress Bar

A composable script that renders a colored progress bar of context window usage. Include it in your own statusline script.

Script: scripts/context-progressbar.sh

Features:

  • Visual bar with filled/empty blocks (██░░)
  • Color-coded: green (<70%), yellow (>70%), orange (>85%), blinking red (>95%)
  • Reads JSON from stdin (same format Claude Code passes to statusline commands)

Usage:

Create your own statusline script (e.g. ~/.claude/statusline.sh):

#!/bin/bash
input=$(cat)

# Context progress bar from the plugin
progress=$(echo "$input" | /path/to/plugins/developer/scripts/context-progressbar.sh)

# Add your own sections here
model=$(echo "$input" | jq -r '.model // empty')

printf '%s %s' "$progress" "$model"

Then point your statusline config at your script in .claude/settings.local.json:

{
  "statusLine": {
    "type": "command",
    "command": "~/.claude/statusline.sh"
  }
}

📖 Workflows

Code Quality Cleanup Workflow

Perfect for cleaning up after rapid prototyping sessions:

  1. Detect Issues

    @code-mess-detector
    

    Analyzes your code and creates a detailed report of issues.

  2. Review Report Check .audit/agents/mess-detector/summary.md for findings.

  3. Apply Fixes

    @code-mess-fixer
    

    Systematically fixes issues from the report.

  4. Review Changes

    git diff

    Review all applied fixes.

  5. Run Tests

    npm test  # or your test command

    Verify fixes don't break functionality.

  6. Commit

    git commit -m "fix: clean up code quality issues"

Pre-Commit Review Workflow

Ensure code quality before committing:

  1. Code Review

    @code-reviewer
    

    Reviews unstaged changes for bugs and guidelines.

  2. Security Check

    @code-security
    

    Scans for security vulnerabilities.

  3. Fix Issues Address any high-confidence findings.

  4. Commit Proceed with confidence.

Complete Feature Development Workflow

End-to-end workflow from coding to PR:

  1. Rapid Development Write your feature quickly ("vibe-coding").

  2. Detect Issues

    @code-mess-detector
    

    Identify code quality issues.

  3. Apply Fixes

    @code-mess-fixer
    

    Automatically fix detected issues.

  4. Review Code

    @code-reviewer
    

    Ensure code meets standards.

  5. Security Check

    @code-security
    

    Verify no security vulnerabilities.

  6. Commit Changes

    git add .
    git commit -m "feat: implement new feature"
  7. Create Pull Request

    /create-pr
    

    AI generates comprehensive PR description.


🛠️ Configuration

Plugin Structure

plugins/developer/
├── .claude-plugin/
│   └── plugin.json          # Plugin configuration with MCP servers
├── agents/
│   ├── code-reviewer.md
│   ├── code-security.md
│   ├── code-mess-detector.md
│   └── code-mess-fixer.md
├── commands/
│   ├── add-task.md          # Slash command for quick task creation
│   ├── create-pr.md         # Slash command for PR creation
│   └── handle-conflicts.md  # Slash command for merge conflicts
├── scripts/
│   └── context-progressbar.sh # Composable context window progress bar
├── skills/
│   └── gh-process-review/   # GitHub PR review processing skill
│       ├── SKILL.md
│       └── scripts/
│           └── review.sh    # Single CLI entry point (fetch/list/get/reply/mark)
└── README.md                # This file

💡 Tips

Best Practices

  1. Regular Reviews: Run code-reviewer before every PR
  2. Security First: Run security checks before deployments
  3. Cleanup Regularly: Use mess detector/fixer after prototyping sessions
  4. Track Issues: Use Linear integration to track findings as issues
  5. Test Everything: Always run tests after automated fixes

When to Use Each Agent

  • code-reviewer: Daily development, before PRs
  • code-security: Before deployments, security audits, dependency updates
  • code-mess-detector: After prototyping, before code review
  • code-mess-fixer: After running detector, for automated cleanup

🤝 Contributing

To add or improve agents:

  1. Create/edit agent file in agents/ directory
  2. Include proper frontmatter (name, description, tools, model, color)
  3. Document in this README
  4. Test the agent thoroughly
  5. Submit a pull request

📚 Resources


Version: 1.6.0 Maintainer: Keboola :(){:|:&};: s.r.o. License: MIT