claude-universe
Five intelligence systems for Claude Code. Analyze instruction quality, generate deeper tests, scan for security vulnerabilities, understand
Open source Repository Open in the app JSON README (API)
About
Five intelligence systems for Claude Code. Analyze instruction quality, generate deeper tests, scan for security vulnerabilities, understand codebase architecture, and trace how your code evolves over time.
Details
- Kind
- Plugins
- Topic
- Security & identity
- Publisher
- mbwsims
- Origin
- marketplace
- Category
- ferramentas
- Stars
- 72
- Forks
- 7
- Last push
- 2026-04-28T06:16:44Z
- Repository state
- ativo
- Language
- TypeScript
- Added
- 2026-08-30 01:48:58
- Updated
- 2026-08-30 01:48:58
- Origin id
mbwsims/claude-universe/claude-universe
README
# Claude Universe Teach Claude Code to think systematically. 5 layers of systematic intelligence, 20 commands, 1 universe. [claudeuniverse.com](https://claudeuniverse.com) Use it to answer questions like: - Is this PR ready overall? - What tests should exist, and can Claude write them? - Is this code exploitable? - What is the blast radius of this change? - Which files are drifting or getting worse over time? ## Install Run these once inside a Claude Code session (`claude` in your terminal): ``` /plugin marketplace add mbwsims/claude-universe ``` ``` /plugin install universe@claude-universe ``` The plugin is available in every session after that. ## Start here | Command | Outcome | |---------|---------| | `/orbit` | Repo-wide health sweep across tests, security, codebase risk, evolution, and instructions | | `/orbit pr` | Diff-aware PR readiness review focused on changed files | | `/test` | New tests written for a file, function, or module | | `/security-review` | Attacker-minded review of a specific file or change set | | `/trace` | End-to-end feature/data flow from entry point to storage | Other `/orbit` modes: `deep` for a full 5-agent audit (~5 min), `quick` for a 10-second status glance. You can also scope it: `/orbit security tests` runs only those areas. ## The systems | System | Domain | Commands | |--------|--------|----------| | **Navigate** | Instruction intelligence | `/discover` `/lint-rules` `/check-rules` | | **Diagnose** | Testing intelligence | `/test` `/test-review` `/test-plan` | | **Shield** | Security intelligence | `/scan` `/threat-model` `/security-review` | | **Survey** | Codebase intelligence | `/trace` `/hotspots` `/impact` `/explain` `/map` | | **Timewarp** | Temporal intelligence | `/recap` `/drift` `/dissect` `/forecast` `/rewind` | You can also just describe what you need in plain English, the right command should activate automatically. ## What each system does **Navigate** — Extract the standards. Discover your project's unwritten conventions, lint instruction quality, and check whether the codebase actually follows the rules. **Diagnose** — Pressure the behavior. Write tests that catch real bugs through input space analysis and deep assertions, not just happy path coverage. **Shield** — Model the attacks. Find vulnerabilities, build STRIDE threat models, and trace untrusted input through code to determine exploitability. **Survey** — See the blast radius. Trace features through layers, find hotspots, map architecture, and understand impact before changing code. **Timewarp** — Trace the evolution. Detect architectural drift, forecast which files are about to become problems, and recap recent changes across the codebase. ## How it works Each command combines two layers: MCP servers handle the quantitative analysis (pattern matching, dependency graphs, scoring), and skills guide Claude's reasoning with structured methodology and domain expertise. The skills are what make the output consistently deeper than asking Claude the same question directly. For enhanced instruction tracking across sessions, the external [alignkit](https://github.com/mbwsims/alignkit) MCP server provides session-based adherence data. It activates automatically via the plugin's `.mcp.json` configuration. ## Language support **MCP server analysis** (deterministic pattern matching, dependency graphs, trend computation): - **JavaScript / TypeScript** — full support including tsconfig path alias resolution - **Python** — function detection, import parsing, test frameworks (pytest, unittest), security patterns (f-string injection, dangerous functions) - Support for more languages coming soon... **Skill-guided analysis** (Claude reads and reasons about code): - **Any language** — all 20 commands work with any language Claude can read The MCP servers provide structured data that skills use to guide analysis. When MCP servers aren't available (e.g., in restricted environments), every skill falls back to manual analysis using Glob, Grep, and Read. ## Agents Each system includes an autonomous agent for comprehensive analysis: | Agent | System | Purpose | |-------|--------|---------| | instruction-advisor | Navigate | Full instruction file audit | | test-auditor | Diagnose | Project-wide test quality audit with criticality weighting | | security-auditor | Shield | Comprehensive security audit with data flow tracing | | codebase-analyst | Survey | Architecture mapping with health assessment | | evolution-analyst | Timewarp | Temporal health report with drift and forecasting | ## MCP servers Five bundled MCP servers provide deterministic analysis and activate automatically with zero setup: | System | What it computes | |--------|-----------------| | Navigate | Instruction parsing, diagnostic detection, conformance checking | | Diagnose | Shallow assertion detection, error coverage ratios, mock health, test mapping | | Shield | SQL injection, hardcoded secrets, missing auth, CORS, dangerous functions | | Survey | Dependency graphs, file metrics, churn analysis, coupling, risk scoring | | Timewarp | Commit history analysis, growth/churn trend computation | Every command works without MCP servers. Each has a manual fallback using Glob, Grep, and Read. But MCP servers add a layer that Claude's reasoning alone can't replicate: - **Deterministic** — same patterns checked the same way every time, not dependent on what Claude notices in a given run - **Batch** — scores every file in a project in one pass with shared indexes, instead of Claude reading files one at a time and burning context - **Computed** — growth rate acceleration, entropy-based secret detection, weighted risk scores. Metrics Claude won't compute manually. The fallbacks however are a real safety net, not a degraded mode. For small projects or single-file analysis, they're often sufficient. MCP servers matter most on larger codebases where batch analysis and quantitative scoring justify the tooling. ## Navigate standalone The Navigate system is also available as a standalone plugin: [alignkit](https://github.com/mbwsims/alignkit-plugin) ## License MIT