\n Slash commands for Claude Code that analyze, validate, and optimize your configuration.\n
\n\n
\n Requires Claude Code CLI — Run these commands inside a Claude Code session\n
\n\n---\n\n> **What is Claude Code?** [Claude Code](https://github.com/anthropics/claude-code) is Anthropic's official agentic coding CLI. Agent Smith adds slash commands that help you check and improve your Claude Code configuration.\n\n---\n\n## What It Does\n\nAgent Smith **analyzes, validates, and fixes** your Claude Code configuration in one interactive workflow:\n\n- **Analyze** — Full 7-pillar evaluation with token metrics, instruction quality, and extension ratings\n- **Track** — Score history with trend indicators and regression detection\n- **Triage** — Interactive walkthrough of findings by effort level\n- **Fix** — Apply accepted changes with an execution plan\n- **Create** — Scaffold new configurations with best practices\n\n---\n\n## Quick Start\n\n```bash\n# Add the marketplace (one-time)\nclaude plugin marketplace add maxencemeloni/claude-code-agent-smith\n\n# Install the plugin\nclaude plugin install agent-smith\n\n# Use (in any project)\ncd your-project\nclaude\n/analyze-agent\n```\n\n> **Update:** `claude plugin marketplace update agent-smith-marketplace && claude plugin update agent-smith@agent-smith-marketplace`\n\n---\n\n## Commands\n\n| Command | Purpose |\n|---------|---------|\n| `/analyze-agent` | Full configuration analysis, interactive triage, and guided fixes |\n| `/create-agent` | Scaffold new configuration with best practices |\n\n### `/analyze-agent` Workflow\n\n1. **Analyze** — Validates structure, scores all 7 pillars, measures tokens, rates instructions and extensions\n2. **Save Report** — Writes full report to `AGENT_SMITH_REPORT.md`\n3. **Triage** — Pick a category: Quick Wins, Recommended, or Advanced\n4. **Decide** — For each finding: Yes (apply) / No (skip) / Custom instruction\n5. **Execute** — Review the execution plan, confirm, and apply all changes\n\n---\n\n## Sample Output\n\n### `/analyze-agent`\n\n```markdown\n# Agent Smith Analysis\n\n**Project:** your-project\n**Type:** Node.js\n**Score:** 7.8/10\n\n## Pillar Scores\n\n| Pillar | Score | Notes |\n|--------|:-----:|-------|\n| Security Posture | 7/10 | Could add more deny rules |\n| Instruction Clarity | 8/10 | Well-structured CLAUDE.md |\n| Configuration Quality | 8/10 | Good structure |\n| Context Efficiency | 6/10 | Missing .claudeignore patterns |\n| Command & Extension Design | 9/10 | Clear, well-documented |\n| Hook Safety | N/A | No hooks configured |\n| MCP Integration | 7/10 | 3 servers configured |\n\n## Content Overview\n\n| Category | Files | Est. Tokens |\n|----------|------:|------------:|\n| Instructions | 1 | ~4,800 |\n| Commands | 5 | ~8,600 |\n| Agents | 6 | ~10,500 |\n| **Total user content** | | **~23,900** |\n\n## Limitations\n\nThis analysis covers user-configurable components only.\nClaude Code's system prompt and tool schemas are outside this scope.\n```\n\n### Interactive Triage\n\n```\nWhich category do you want to address?\n A) Quick Wins — 3 low-effort fixes\n B) Recommended — 2 medium-effort improvements\n C) Advanced — 1 high-effort optimization\n D) All categories — Walk through everything\n E) Done — Just the report, no changes\n```\n\nAfter choosing, you walk through each item one by one and decide what to do. Then Agent Smith builds an execution plan and applies your choices.\n\n> See [`examples/reports/`](./examples/reports/) for a full real-world report from Agent Smith's self-analysis, and [`examples/configs/`](./examples/configs/) for starter configurations for Node.js, Python, and generic projects.\n\n---\n\n## The 7 Evaluation Areas\n\n| Area | Weight | What's Checked |\n|------|:------:|----------------|\n| **Security Posture** | 20% | Sensitive file protection, dangerous patterns |\n| **Instruction Clarity** | 20% | CLAUDE.md quality, structure, contradictions |\n| **Configuration Quality** | 15% | settings.json structure, allow rules |\n| **Context Efficiency** | 15% | .claudeignore coverage, duplication, references |\n| **Command & Extension Design** | 15% | Commands, agents, skills: quality, naming, structure |\n| **Hook Safety** | 10% | hooks.json validity, dangerous commands |\n| **MCP Integration** | 5% | MCP server configuration quality |\n\n---\n\n## What Issues It Finds\n\n| Issue Type | Example |\n|------------|---------|\n| **Missing .claudeignore** | No context filtering → node_modules scanned |\n| **Missing .git/ pattern** | Git history gets indexed |\n| **Missing deny rules** | No protection for .env, secrets/, *.pem |\n| **Dangerous Bash patterns** | `Bash(*)` allows any command |\n| **Unsafe hook commands** | `rm -rf` or `sudo` in hooks |\n| **Hardcoded personal paths** | `/Users/name/` in shared configs |\n| **--no-verify bypasses** | Git safety hooks disabled |\n| **Unscoped agents** | Missing model or tools in agent frontmatter |\n| **Unstructured skills** | Missing SKILL.md or frontmatter |\n| **Too many active MCPs** | >10 servers shrinks context window |\n| **Duplicated content** | Same text in multiple files |\n| **Embedded content** | File contents copied instead of referenced |\n\n---\n\n## Honest About Limitations\n\nThis tool measures what you can control:\n\n| Can Measure | Cannot Measure |\n|-------------|----------------|\n| Your instruction files | Claude Code's system prompt |\n| Your commands and agents | Built-in tool schemas |\n| Your .claudeignore coverage | Runtime context usage |\n| Duplicated content | Conversation history |\n\nWe provide **real measurements** with **honest scope**, not inflated claims.\n\n---\n\n## Security: Deny Rules\n\nPrevent Claude from reading sensitive files:\n\n```json\n{\n \"permissions\": {\n \"deny\": [\n \"Read(./.env)\",\n \"Read(./.env.*)\",\n \"Read(./secrets/)\",\n \"Read(./**/*.pem)\",\n \"Read(./**/*.key)\",\n \"Read(./**/*_rsa)\"\n ]\n }\n}\n```\n\n`/create-agent` includes these by default. `/analyze-agent` checks for missing rules.\n\n---\n\n## Install\n\n```bash\n# 1. Add the marketplace (one-time)\nclaude plugin marketplace add maxencemeloni/claude-code-agent-smith\n\n# 2. Install the plugin\nclaude plugin install agent-smith\n```\n\nWorks on all platforms.\n\n```bash\n# Update (refresh marketplace first, then update)\nclaude plugin marketplace update agent-smith-marketplace\nclaude plugin update agent-smith@agent-smith-marketplace\n\n# Uninstall\nclaude plugin uninstall agent-smith@agent-smith-marketplace\n```\n\n> **Upgrading from v1.x?** The install script (`install.sh`/`install.ps1`) method was removed in v2.0. To migrate:\n> 1. Run the old uninstall script: `bash uninstall.sh` (or `.\\uninstall.ps1` on Windows)\n> 2. Install fresh using the plugin commands above\n>\n> If you no longer have the uninstall script, manually delete the Agent Smith commands from `~/.claude/commands/` and the `~/.claude/agent-smith-repo` file, then install the plugin.\n\n---\n\n## Contributing\n\n1. Fork the repo\n2. Add/modify commands in `commands/`\n3. Validate with `claude plugin validate .`\n4. Submit a PR\n\nSee [IMPROVEMENTS.md](./IMPROVEMENTS.md) for the full changelog and the [wiki Roadmap](https://github.com/maxencemeloni/claude-code-agent-smith/wiki/Roadmap) for planned features.\n\n---\n\n## Development\n\nWhen working on Agent Smith with Claude Code, the `CLAUDE.md` file at the project root provides development context including:\n\n- **Design principles** — Honesty over hype, scope clarity, evidence required\n- **7-pillar system** — Rationale for each evaluation area\n- **Out of scope** — What we explicitly don't measure (system prompt, model routing, etc.)\n- **Version management** — How to bump versions and update documentation\n- **Repository structure** — Where everything lives\n\nThis context is only loaded when developing Agent Smith itself, not when users run the commands in their projects.\n\n### Release Checklist\n\nWhen releasing a new version:\n\n| Step | Files |\n|------|-------|\n| 1. Update version badge | `README.md` |\n| 2. Add changelog entry | `IMPROVEMENTS.md` |\n| 3. Update wiki | `Roadmap.md` (version + history) |\n| 4. Commit with format | `Release vX.Y.Z — Description` |\n| 5. Push main + wiki | Both repositories |\n\n### Wiki\n\nThe [GitHub Wiki](https://github.com/maxencemeloni/claude-code-agent-smith/wiki) is maintained as a separate repository. When making changes that affect documentation, update both the README and relevant wiki pages.\n\n---\n\n## License\n\nMIT\n\n---\n\n## More Claude Code Plugins\n\nTools built by the same developer to enhance your Claude Code experience.\n\n| Plugin | Description | Command |\n|--------|-------------|---------|\n| **[Prompt Smith](https://github.com/maxencemeloni/claude-code-prompt-smith)** | Optimizes rough prompts into cleaner, more executable versions. Four modes: default, agentic, compact, and strict. | `/prompt` |\n\n> Check out [Prompt Smith](https://prompt-smith.mmapi.fr/) for more details\n\n---\n\n
![\"Claude](\"./assets/banner.png\")
\n![\"Claude](\"https://img.shields.io/badge/Claude_Code-Slash_Commands-blueviolet?style=for-the-badge\"){kind=icon}
![\"Version\"/](\"https://img.shields.io/badge/v2.5.1-Stable-green?style=for-the-badge\"){kind=icon}
![\"License\"/](\"https://img.shields.io/badge/MIT-License-blue?style=for-the-badge\"){kind=icon}