{
  "markdown": "# Claude Universe\n\nTeach Claude Code to think systematically.\n\n5 layers of systematic intelligence, 20 commands, 1 universe.\n\n[claudeuniverse.com](https://claudeuniverse.com)\n\nUse it to answer questions like:\n\n- Is this PR ready overall?\n- What tests should exist, and can Claude write them?\n- Is this code exploitable?\n- What is the blast radius of this change?\n- Which files are drifting or getting worse over time?\n\n\n## Install\n\nRun these once inside a Claude Code session (`claude` in your terminal):\n\n```\n/plugin marketplace add mbwsims/claude-universe\n```\n\n```\n/plugin install universe@claude-universe\n```\n\nThe plugin is available in every session after that.\n\n\n## Start here\n\n| Command | Outcome |\n|---------|---------|\n| `/orbit` | Repo-wide health sweep across tests, security, codebase risk, evolution, and instructions |\n| `/orbit pr` | Diff-aware PR readiness review focused on changed files |\n| `/test` | New tests written for a file, function, or module |\n| `/security-review` | Attacker-minded review of a specific file or change set |\n| `/trace` | End-to-end feature/data flow from entry point to storage |\n\nOther `/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.\n\n\n## The systems\n\n| System | Domain | Commands |\n|--------|--------|----------|\n| **Navigate** | Instruction intelligence | `/discover` `/lint-rules` `/check-rules` |\n| **Diagnose** | Testing intelligence | `/test` `/test-review` `/test-plan` |\n| **Shield** | Security intelligence | `/scan` `/threat-model` `/security-review` |\n| **Survey** | Codebase intelligence | `/trace` `/hotspots` `/impact` `/explain` `/map` |\n| **Timewarp** | Temporal intelligence | `/recap` `/drift` `/dissect` `/forecast` `/rewind` |\n\nYou can also just describe what you need in plain English, the right command should activate automatically.\n\n## What each system does\n\n**Navigate** — Extract the standards. Discover your project's unwritten conventions,\nlint instruction quality, and check whether the codebase actually follows the rules.\n\n**Diagnose** — Pressure the behavior. Write tests that catch real bugs through input\nspace analysis and deep assertions, not just happy path coverage.\n\n**Shield** — Model the attacks. Find vulnerabilities, build STRIDE threat models,\nand trace untrusted input through code to determine exploitability.\n\n**Survey** — See the blast radius. Trace features through layers, find hotspots,\nmap architecture, and understand impact before changing code.\n\n**Timewarp** — Trace the evolution. Detect architectural drift, forecast which files\nare about to become problems, and recap recent changes across the codebase.\n\n\n## How it works\n\nEach command combines two layers: MCP servers handle the quantitative\nanalysis (pattern matching, dependency graphs, scoring), and skills guide\nClaude's reasoning with structured methodology and domain expertise. The\nskills are what make the output consistently deeper than asking Claude\nthe same question directly.\n\nFor enhanced instruction tracking across sessions, the external [alignkit](https://github.com/mbwsims/alignkit) MCP\nserver provides session-based adherence data. It activates automatically via the\nplugin's `.mcp.json` configuration.\n\n\n## Language support\n\n**MCP server analysis** (deterministic pattern matching, dependency graphs, trend computation):\n- **JavaScript / TypeScript** — full support including tsconfig path alias resolution\n- **Python** — function detection, import parsing, test frameworks (pytest, unittest), security patterns (f-string injection, dangerous functions)\n- Support for more languages coming soon...\n\n**Skill-guided analysis** (Claude reads and reasons about code):\n- **Any language** — all 20 commands work with any language Claude can read\n\nThe MCP servers provide structured data that skills use to guide analysis. When MCP\nservers aren't available (e.g., in restricted environments), every skill falls back\nto manual analysis using Glob, Grep, and Read.\n\n\n## Agents\n\nEach system includes an autonomous agent for comprehensive analysis:\n\n| Agent | System | Purpose |\n|-------|--------|---------|\n| instruction-advisor | Navigate | Full instruction file audit |\n| test-auditor | Diagnose | Project-wide test quality audit with criticality weighting |\n| security-auditor | Shield | Comprehensive security audit with data flow tracing |\n| codebase-analyst | Survey | Architecture mapping with health assessment |\n| evolution-analyst | Timewarp | Temporal health report with drift and forecasting |\n\n\n## MCP servers\n\nFive bundled MCP servers provide deterministic analysis and activate automatically\nwith zero setup:\n\n| System | What it computes |\n|--------|-----------------|\n| Navigate | Instruction parsing, diagnostic detection, conformance checking |\n| Diagnose | Shallow assertion detection, error coverage ratios, mock health, test mapping |\n| Shield | SQL injection, hardcoded secrets, missing auth, CORS, dangerous functions |\n| Survey | Dependency graphs, file metrics, churn analysis, coupling, risk scoring |\n| Timewarp | Commit history analysis, growth/churn trend computation |\n\nEvery command works without MCP servers. Each has a manual fallback using Glob,\nGrep, and Read. But MCP servers add a layer that Claude's reasoning alone can't\nreplicate:\n\n- **Deterministic** — same patterns checked the same way every time, not dependent\n  on what Claude notices in a given run\n- **Batch** — scores every file in a project in one pass with shared indexes, instead\n  of Claude reading files one at a time and burning context\n- **Computed** — growth rate acceleration, entropy-based secret detection, weighted\n  risk scores. Metrics Claude won't compute manually.\n\nThe fallbacks however are a real safety net, not a degraded mode. For small projects or\nsingle-file analysis, they're often sufficient. MCP servers matter most on larger\ncodebases where batch analysis and quantitative scoring justify the tooling.\n\n\n## Navigate standalone\n\nThe Navigate system is also available as a standalone plugin:\n[alignkit](https://github.com/mbwsims/alignkit-plugin)\n\n\n## License\n\nMIT\n",
  "bytes": 6170,
  "sha": "f2fb6da4a9a78fee4ee3d92c7f7a069851fab33844cea5b215121098bfbbbf00",
  "repo_slug": "mbwsims/claude-universe",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_mbwsims_claude_universe_claude_universe_308a5194/readme"
}