Sub-agent: invoked only by the orchestrator-agent for bug intents. Reproduces the bug with a failing test, isolates the root cause, applies a minimal patch, and verifies all tests pass. Scoped strictly to the broken behavior — no refactors, no opportunistic cleanup. Do not invoke directly.
Sub-agent: invoked only by the orchestrator-agent when a design_system intent is detected. Runs design-expert → design-system-docs → plan-expert in sequence to document design tokens, audit or plan Storybook, and create an execution plan. Do not invoke directly.
Sub-agent: invoked only by the orchestrator-agent when new_feature or planning-features-agent flow is active. Conducts structured discovery interviews to transform vague ideas into precise feature specifications and ClickUp tickets. Do not invoke directly.
Sub-agent: invoked only by the orchestrator-agent to execute a specific subtask from plan to Pull Request. Reads project context, writes production-ready code, runs code-review and a11y-auditor, commits, and opens a PR via create-pr. Do not invoke directly.
The default entry point for ai-toolbox. Use this agent for ANY user request — feature planning, task implementation, code review, design systems, accessibility, or knowledge management. Analyzes intent and routes to the correct sub-agent automatically. Checks for WIKI.md at startup and delegates to wiki-agent to initialize the wiki if it is missing before any other step. Examples: "I want to plan a new feature", "Implement ticket CU-abc123", "Set up the design system for this project", "Review my code changes before I commit", "Work on this task and open a PR when done".
Audits code, components, or screenshots for accessibility barriers following WCAG 2.2 (Levels A, AA, AAA). Detects the environment automatically and runs only the relevant checks — Web (HTML, React, Next.js, Tailwind) or Mobile (React Native, Expo, Swift, Kotlin).
Two-phase code review. Phase 1 runs fast pre-commit checks for spec compliance, type safety, and stack alignment. Phase 2 audits SOLID principles and structural integrity. Use before committing or opening a PR.
Acts as an issue triager to gather all required information for a bug or issue, creates a ticket in ClickUp, and then creates an issue in GitHub natively using the templates.
**Role:** Senior engineer opening a pull request.
Creates a ClickUp task from raw input (client message, rough idea, or transcribed audio) by classifying it into one of 5 task types ([US], [BUG], [IMP], [TASK], [SPIKE]), filling out the standardized template, and submitting it to ClickUp.
Executes bash commands
Hook triggers when Bash tool is used
Modifies files
Hook triggers on file write and edit operations
External network access
Connects to servers outside your machine
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
A Claude Code plugin with a curated set of skills and agents for software teams. Covers accessibility auditing, code review, project initialization, design system documentation, and end-to-end feature planning workflows backed by ClickUp.
Skills are reusable workflows invoked with a / command directly in Claude Code. Each skill is independent and can be used on its own.
| Skill | Command | What it does |
|---|---|---|
| init-project | /init-project | Scans the codebase and generates an AGENTS.md with the detected stack, structure, and dev commands. Run this first on any new project. |
| code-review | /code-review | Two-phase code review: fast pre-commit checks (spec compliance, type safety, security) followed by a deep SOLID / KISS / DRY structural audit. |
| a11y-auditor | /a11y-auditor | Audits code or components for accessibility barriers against WCAG 2.2 (A, AA, AAA). Auto-detects web vs. mobile stack. |
| feature-discovery | /feature-discovery | Acts as a functional analyst to gather all feature requirements through structured questioning. Outputs a comprehensive spec and optionally creates a ClickUp ticket. |
| plan-expert | /plan-expert | Takes a ClickUp ticket or a free-form description and breaks it into detailed, ordered subtasks using a structured 8-section template. Creates subtasks on the ClickUp ticket or as a local task list. |
| design-expert | /design-expert | Scans the project for all design-related information (colors, typography, spacing, component patterns, dark mode, design system) and generates or updates a DESIGN.md file. |
| design-system-docs | /design-system-docs | Audits design system documentation. If Storybook is present, reviews its quality and suggests improvements. If not, produces a step-by-step plan to implement it. |
| design-system-setup | /design-system-setup | End-to-end design system setup. Runs design-expert → design-system-docs → plan-expert in sequence to document the design system, audit or plan Storybook, and create all execution tasks in ClickUp or locally. |
| planning-features | /planning-features | End-to-end feature planning. Runs feature-discovery then plan-expert back to back — gathers requirements, creates a ClickUp ticket, and breaks it into an execution plan. |
| create-pr | /create-pr | Creates a GitHub PR with a fully auto-populated standardized template. Infers base branch, derives description from the diff, detects shared code impact, tags stakeholders from CODEOWNERS, and builds a concrete test plan. Designed to run without human input when called by an agent. |
| implement-task | /implement-task | Implements a task end-to-end. Given a ClickUp ticket ID or description, reads project context, plans at the file level, writes the code, runs automated checks + code-review, applies fixes, commits, and opens a PR via create-pr. |
Agents follow an orchestrator → sub-agent architecture. The orchestrator-agent is the only agent Claude auto-selects — it analyzes every user request and routes to the correct specialist sub-agent.
| Agent | Role |
|---|---|
| orchestrator-agent | Default entry point. Handles any user request by classifying intent and routing to the right sub-agent. Always runs first; always responds last. |
| Agent | Activated when |
|---|---|
| planning-features-agent | new_feature intent — coordinates discovery + planning end-to-end. |
| feature-discovery-agent | Called by planning-features-agent — structured requirement interviews → FEATURE_SPEC + ClickUp ticket. |
| plan-expert-agent | quick_task, refactor, or after discovery — decomposes specs into 8-section subtasks. |
| implement-task-agent | implementation intent or after planning — writes code, runs review, commits, opens PR. |
| design-system-setup-agent | design_system intent — design-expert → design-system-docs → plan-expert pipeline. |
How to use: Just describe what you want in natural language. The orchestrator routes automatically. Use
/skills for direct, one-off invocations when you know exactly which step to run.
Run this from your terminal — no cloning or manual steps required:
curl -fsSL https://raw.githubusercontent.com/Axis-Human/dev-workflow-plugin/main/install.sh | bash
This installs the plugin and configures the required hooks in ~/.claude/settings.json automatically.
npx claudepluginhub axis-human/dev-workflow-plugin --plugin axis-human-ai-toolboxUltra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Harness-native ECC plugin for engineering teams - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Memory compression system for Claude Code - persist context across sessions
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques