A comprehensive toolkit for developers including specialized agents for code review, security analysis, and code quality management.
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
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
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
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
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 handlingCommand: /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 draftPrerequisites:
- GitHub CLI (
gh) installed and authenticated - Branch has commits not in base branch
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-conflictsPrerequisites:
- Active merge/rebase/cherry-pick with conflicts
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 continueWorkflow:
- Fetches all review threads to local JSON
- Lists unresolved threads needing attention
- For each thread: read feedback → implement fix → commit → push → reply with commit hash
- Marks threads as locally resolved for tracking
Prerequisites:
- GitHub CLI (
gh) installed and authenticated - Current branch has an associated PR with review comments
jqinstalled for JSON processing
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:
- Plugin automatically configures the Linear MCP
- Run
/mcpto authenticate with Linear (OAuth) - Start using Linear tools in your workflow
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"
}
}Perfect for cleaning up after rapid prototyping sessions:
-
Detect Issues
@code-mess-detectorAnalyzes your code and creates a detailed report of issues.
-
Review Report Check
.audit/agents/mess-detector/summary.mdfor findings. -
Apply Fixes
@code-mess-fixerSystematically fixes issues from the report.
-
Review Changes
git diff
Review all applied fixes.
-
Run Tests
npm test # or your test command
Verify fixes don't break functionality.
-
Commit
git commit -m "fix: clean up code quality issues"
Ensure code quality before committing:
-
Code Review
@code-reviewerReviews unstaged changes for bugs and guidelines.
-
Security Check
@code-securityScans for security vulnerabilities.
-
Fix Issues Address any high-confidence findings.
-
Commit Proceed with confidence.
End-to-end workflow from coding to PR:
-
Rapid Development Write your feature quickly ("vibe-coding").
-
Detect Issues
@code-mess-detectorIdentify code quality issues.
-
Apply Fixes
@code-mess-fixerAutomatically fix detected issues.
-
Review Code
@code-reviewerEnsure code meets standards.
-
Security Check
@code-securityVerify no security vulnerabilities.
-
Commit Changes
git add . git commit -m "feat: implement new feature"
-
Create Pull Request
/create-prAI generates comprehensive PR description.
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
- Regular Reviews: Run code-reviewer before every PR
- Security First: Run security checks before deployments
- Cleanup Regularly: Use mess detector/fixer after prototyping sessions
- Track Issues: Use Linear integration to track findings as issues
- Test Everything: Always run tests after automated fixes
- 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
To add or improve agents:
- Create/edit agent file in
agents/directory - Include proper frontmatter (name, description, tools, model, color)
- Document in this README
- Test the agent thoroughly
- Submit a pull request
Version: 1.6.0 Maintainer: Keboola :(){:|:&};: s.r.o. License: MIT