Turn Claude into your personal development team. Plugin skills that handle everything — from deep planning through implementation to code review, with parallel agent swarms and automatic quality gates.
v7.5.0 —
/setup-swarm: One-Command Agent Teams SetupNew
/setup-swarmskill enables the Agent Teams feature in yoursettings.jsonautomatically — no manual JSON editing needed. Required for/hydraand/legion. Both swarm skills now warn you to close all other Claude Code sessions before setup, since editingsettings.jsoncan crash active sessions.The Claude Agent System is distributed exclusively as a Claude Code plugin. If you previously installed via the legacy setup script, uninstall the old files first:
rm -rf .claude/commands .claude/workflows .claude/middleware .claude/agents /plugin marketplace add Kasempiternal/Claude-Agent-System /plugin install cas
/plugin marketplace add Kasempiternal/Claude-Agent-System
/plugin install casDone! You now have 9 skills: /zk, /legion, /pcc, /pcc-opus, /hydra, /review, /systemcc, /setup-swarm, and /setup-hooks.
"Vibe coding" = typing a prompt, accepting whatever the AI outputs, hoping it works.
This system is the opposite. Every skill enforces structure:
| What Vibe Coding Does | What This System Does |
|---|---|
| Blindly accepts AI output | Parallel code review agents |
| No validation | Build config detection + linting enforcement |
| No learning | Session memory - learns your patterns and mistakes |
| No quality gates | Decision engines with complexity/risk/scope analysis |
| Hope it works | Post-execution validation + auto-fix critical issues |
The smart entry point. Analyzes your request and auto-routes to the best execution mode — no manual skill selection needed.
/zk "what you want to do"ZK walks a deterministic 5-step decision tree:
| Step | Condition | Routes To |
|---|---|---|
| 0 | Holistic project needing iterative completion? | /legion |
| 1 | Multiple independent deliverables? | /hydra |
| 2 | Scale word + broad noun ("entire codebase")? | /hydra |
| 3 | High-stakes keyword + qualifying signal? | /pcc-opus |
| 4 | Everything else (default) | /pcc |
/zk build a complete todo app from scratch # -> Legion (holistic project, iterative)
/zk add a button to the settings page # -> PCC (simple, clear scope)
/zk refactor the payment processing system # -> PCC-Opus (keyword + risk domain)
/zk migrate all models to SwiftData # -> PCC-Opus ("migrate" always qualifies)
/zk fix auth; add dashboard; update API # -> Hydra (3 independent tasks)
/zk modernize the entire codebase # -> Hydra (scale + broad scope)Escape hatch: You can always bypass ZK and invoke /legion, /pcc, /pcc-opus, or /hydra directly.
An orchestrator that spawns agent swarms for exploration and implementation.
/pcc "implement user authentication with JWT tokens"
/pcc-opus "refactor the entire payment processing system"| Variant | Scouts | Implementers | Best For |
|---|---|---|---|
/pcc |
Sonnet (fast, cost-efficient) | Opus (high quality) | Most tasks |
/pcc-opus |
Opus (maximum depth) | Opus (high quality) | Critical systems, unfamiliar codebases |
- Task Understanding - Clarifies the task with you
- Parallel Exploration - Spawns 2-6 scout agents to map the codebase
- Synthesis - Combines findings into unified understanding
- Clarification - Asks questions if multiple valid approaches exist
- Plan Creation - Creates editable plan at
.claude/plans/{task}.md - User Review - You edit and approve the plan before any code is written
- Parallel Implementation - Spawns 2-6 Opus agents working simultaneously
- Verification - Tests and code review
- Simplification - 2-6 parallel agents clean up the code
- Final Report - Summarizes everything
Submit a holistic project description. Legion deploys a full agent swarm each iteration — scouts, CTO analyst, wave-based implementers, verifiers — then checks if the project is complete. It keeps iterating autonomously until everything is built, the max iteration limit is hit, or progress stalls.
Requires Agent Teams: Run
/setup-swarmto enable this automatically.⚠️ Close all other Claude Code sessions first — editingsettings.jsonwhile other sessions run can crash them.
Very High Token Usage Warning: Legion runs multiple iterations of agent swarms. Each iteration spawns 5-30 Opus agents. Recommended only for MAX plan subscribers.
/legion build a complete todo app with local storage from scratch
/legion create an e-commerce platform with auth, cart, and checkout --max-iterations 8
/legion implement the full API layer end to end --checkpoint- Building complete features or applications from scratch
- Projects that need multiple rounds of build-test-fix
- When you want autonomous completion without manual re-runs
- Prerequisites Check - Verifies
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 - Project Parse - Parses holistic project description, extracts
--max-iterationsand--checkpointflags - Team Init - Creates team + structural phase tasks
- Full Exploration - Opus scout teammates explore the entire project scope
- CTO Analysis - CTO analyst creates master task list, decomposes project into modules and waves
- User Confirmation - You review the CTO's plan, then confirm
- Iteration Loop - The core autonomous loop:
- Iteration 1: full wave-based implementation (Hydra-scale)
- Iteration 2+: delta scouts -> CTO updates task list -> targeted implementation
- Each iteration: verify -> assess completion -> loop or exit
- Exit: all P1 tasks done + tests pass, OR max iterations, OR stall detected
- Simplification - Module-grouped cleanup (only if project completed)
- Final Report & Cleanup - Iteration log, final task status, shutdown, clean up
- Autonomous iteration loop — keeps deploying swarms until the project is done
- Master task list — living checkbox document, updated each iteration by the CTO analyst
- Risk tier classification — every task gets a tier (T0-T3) with tier-scaled verification depth
- Recovery procedures — stuck agent replacement, partial rollback, budget overrun protection, context pressure conservation
- Anti-pattern validation — catches coordinator-implements-code, redundant agents, file overlap, scope drift
- Iteration scaling — iteration 1 is heavy (15-30 agents), iteration 2+ is light (5-12 agents)
- Circuit breaker — stops after 2 consecutive iterations with no progress
- Checkpoint mode (
--checkpoint) — optional pause between iterations for user approval - Configurable max iterations (
--max-iterations N, default 5) - Post-loop simplification — module-grouped code cleanup after project completion
Submit N tasks at once. Hydra plans them together, detects cross-task file conflicts, then deploys implementation swarms in dependency-ordered waves — fully parallel where files don't overlap, sequentially ordered where they do.
Requires Agent Teams: Run
/setup-swarmto enable this automatically.⚠️ Close all other Claude Code sessions first — editingsettings.jsonwhile other sessions run can crash them.
High Token Usage Warning: Hydra spawns multiple Opus-level agents in parallel swarms, which can result in very high token consumption. Recommended only for MAX plan subscribers.
/hydra add auth; build dashboard; fix payments
/hydra 1. refactor models 2. add API endpoints 3. update tests- Multiple independent tasks in the same project
- Batch implementation sprints
- When tasks might share files (Hydra detects and resolves conflicts)
- Prerequisites Check - Verifies
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 - Task Parsing - Splits input into N discrete tasks
- Team Init - Creates team + tasks for all N items
- Parallel Exploration - Shared pool of Opus scout teammates explores for all N tasks
- Delegated Synthesis - Analyst teammate writes N plans + coordination file, resolves conflicts
- User Review - You review summary + plan files, then confirm
- Wave Implementation - Per wave: analyst prepares specs, orchestrator spawns impl agents
- Per-Wave Verification - Tests after each wave; global integration check after all waves
- Simplification - Module-grouped cleanup across all task boundaries
- Final Report & Cleanup - Per-task status, shutdown teammates, clean up
- Agent Teams powered — structured coordination with TeamCreate, TaskCreate, SendMessage
- Cross-task file conflict analysis — builds a DAG of file ownership at plan time
- Wave-based execution — parallel where safe, sequential where files overlap
- Risk tier classification — every task gets a tier (T0-T3) with tier-scaled verification depth
- Recovery procedures — stuck agent replacement and partial rollback on verification failure
- Anti-pattern validation — catches redundant agents, file overlap, sequential deps in same wave
- Shared scout pool — Opus scouts explore for all N tasks simultaneously
- Per-task plans + coordination file — editable before implementation
- Module-grouped simplification — ensures cross-task code consistency
Deploys 7 parallel review agents to analyze your code, then automatically fixes CRITICAL and MAJOR findings with your approval. Uses official Anthropic review plugin agents when available, with bundled fallback agents for standalone use.
/review # Review all uncommitted changes (default)
/review staged # Review only staged changes
/review src/auth.ts # Review specific file(s)
/review "auth module" # Review files matching a descriptionAll 7 agents run in parallel (same wall-clock time as running one):
| Agent | What It Checks |
|---|---|
| Bug & Logic Reviewer | Security vulnerabilities, crashes, logic errors, resource leaks |
| Code Reviewer | Code quality, readability, maintainability, best practices |
| Project Guidelines Reviewer | Style conventions, CLAUDE.md standards, best practices |
| Silent Failure Hunter | Swallowed exceptions, bad fallbacks, inadequate error handling |
| Comment Analyzer | Stale docs, misleading comments, missing documentation |
| Type Design Analyzer | Encapsulation, invariant expression, type safety |
| Test Coverage Analyzer | Test gaps, missing edge cases, test quality |
The orchestrator synthesizes all 7 agent reports into a consolidated review:
- Health score (0-10) with severity-weighted formula
- Agent verdicts table - quick pass/fail per agent
- Deduplicated findings - overlapping issues merged, multi-agent flags boost confidence
- Cross-agent correlation - related findings from different agents grouped together
- Severity-prioritized - CRITICAL > MAJOR > MINOR
If CRITICAL or MAJOR findings are found, the system asks how you want to proceed:
- Fix CRITICAL and MAJOR (default) - parallel fix agents resolve high-severity findings
- Fix ALL - also addresses MINOR findings (style, comments, naming)
- Report only - keep the report without modifying code
Fix agents are grouped by file (exclusive ownership, no conflicts) and make minimum changes to resolve each finding. /review never modifies code without your explicit consent.
/review works standalone with bundled agents, but installing these official plugins enhances the analysis:
/plugin install pr-review-toolkit@claude-plugins-official
/plugin install code-simplifier@claude-plugins-officialThe catch-all convenience command. Auto-analyzes task complexity, risk, and scope to select and execute the optimal workflow automatically.
/systemcc "what you want to do"- Lyra AI Optimization - Enhances your request with the 4-D methodology
- Build Config Detection - Scans Makefile, CI/CD, linters and applies rules
- Two-Phase Decision Engine - Domain detection first, then complexity scoring
- Automatic Execution - Runs all phases without manual intervention
- Triple Code Review - 3 parallel reviewers check quality
- Summary - What changed and why
- 3-dimensional task analysis (complexity, risk, scope)
- Two-phase decision engine with confidence scoring
- Lyra AI prompt optimization (4-D methodology)
- Build configuration auto-detection and enforcement
- Automatic workflow selection and execution
- Triple code review (Senior Engineer, Lead Engineer, Architect)
Enables the CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS env var in your ~/.claude/settings.json. Required before using /hydra or /legion.
/setup-swarm
⚠️ Close all other Claude Code sessions first — editingsettings.jsonwhile other sessions are running can crash or corrupt those sessions. The skill will warn you about this before making changes.
- Run once — the setting persists across all projects and sessions
- Non-destructive — merges into existing settings without overwriting
- Restart required — Claude Code needs a restart after the change
Installs PreToolUse hooks into your ~/.claude/settings.json that intercept risky actions and prompt you before proceeding.
/setup-hooks| Hook | What It Catches |
|---|---|
| push-guard | Any git push or gh pr create — commits are allowed freely |
| dangerous-commands | rm -rf ~/, dd to disk, git reset --hard, curl | sh, fork bombs, etc. |
| protect-secrets | .env files, SSH keys, AWS creds, secret variables, exfiltration attempts |
All hooks use "ask" mode — Claude pauses and shows you a yes/no prompt instead of silently blocking. Approve when you asked for the action, deny when Claude acts autonomously.
- Run once — hooks persist in your settings across all projects and sessions
- Selective install — choose which hooks you want during setup
- Non-destructive — merges into existing settings without overwriting
- Audit logging — all intercepted actions logged to
~/.claude/hooks-logs/
Based on karanb192/claude-code-hooks, modified to use
"ask"instead of"deny".
| Situation | Use This |
|---|---|
| Don't want to choose — let the system pick | /zk |
| Build a complete project from scratch | /legion |
| Single well-defined task | /pcc |
| Critical systems, unfamiliar codebases | /pcc-opus |
| Multiple independent tasks at once | /hydra (or /zk auto-detects) |
| Code review before committing | /review |
| Want auto-routing with Lyra AI optimization | /systemcc |
| Enable Agent Teams for Hydra/Legion | /setup-swarm (run once) |
| Prevent Claude from pushing without permission | /setup-hooks (run once) |
Contributions are welcome! See CONTRIBUTING.md for guidelines.
The /systemcc skill was partially inspired by ideas shared in the community:
- Anti-YOLO Method - ASCII wireframing for web projects
- Phase-based development - Large codebase handling
- Multi-agent workflows - Team-based development
- Agent OS - Project initialization framework
All other skills (/zk, /legion, /pcc, /pcc-opus, /hydra, /review) are original.
MIT License - see LICENSE for details.