# TechRex Claude Code Setup Bundle > Drop-in config bundle for Claude Code. Hand the prompt below to any > AI coding agent (Claude Code, Cursor, Codex) together with the public > repo URL, and it will install the full TechRex setup onto your machine. > > - Repo: https://github.com/saiteja007-mv/techrex-claude-setup > - Targets: `~/.claude/` (Linux/macOS) or `%USERPROFILE%\.claude` (Windows) > - Channel: https://www.youtube.com/@The_TechRex --- ## What gets installed | Category | Count | What it is | |-----------------|-------|------------------------------------------------------------| | Global CLAUDE.md| 1 | Caveman Mode + RTK (Rust Token Killer) + graphify reference| | Plugins | 11 | Marketplace plugins installed via `/plugin install` | | Custom skills | 17 | autoresearch, graphify, data-analyst, seedance, ... | | Sub-agents | 6 | blog-writer, QA, fullstack, system-troubleshooter, ... | | Slash commands | 29 | `gsd/*` (26) + `vibe-*` (3) | | Hook | 1 | PreToolUse Glob\|Grep -> graphify reminder | | MCP stubs | 2 | `search-console` + `n8n-mcp` templates (no secrets) | | Statusline | opt-in| ccstatusline launcher | Everything is scrubbed — no API keys, no JWTs, no OAuth tokens, no ngrok URLs, no email addresses. You configure your own after install. --- ## Part 1 — the agent prompt Copy the whole block and paste into your AI agent. The site also hosts the plain-text version at [`/claude-code-setup/agent-prompt.txt`](/claude-code-setup/agent-prompt.txt) if you want to curl it. ``` You are installing the TechRex Claude Code Setup Bundle onto this machine. Repo: https://github.com/saiteja007-mv/techrex-claude-setup Target: the current user's ~/.claude/ directory. Rules of engagement: - NEVER overwrite existing files I didn't tell you to overwrite. - NEVER clobber existing settings.json / config.json — merge only. - ALWAYS back up the existing ~/.claude before touching it. - After each step, print ONE line: "v step N — ". - At the end, print a table: v / ! / x per section. Run these steps in order: 1. DETECT ENVIRONMENT — OS, $HOME, confirm `claude` CLI on PATH. 2. BACKUP — copy ~/.claude to ~/.claude.bak-/. 3. CLONE — git clone https://github.com/saiteja007-mv/techrex-claude-setup into a temp dir (git pull if already present). 4. MERGE CLAUDE.md — if mine doesn't exist, copy verbatim; else append the repo's content between markers: # === BEGIN TechRex Bundle === ... # === END TechRex Bundle === 5. COPY RULES — copy .claude/rules/*.md if not present (non-destructive). 6. COPY SKILLS — copy each dir in .claude/skills/ if not present. 7. COPY AGENTS — copy each .md in .claude/agents/ if absent. 8. COPY COMMANDS — copy every .md in .claude/commands/ if absent. 9. STATUSLINE — ask "Install TechRex statusline? [y/N]"; if yes, copy statusline.bat + statusline.js. 10. SETTINGS.JSON — load .claude/settings.json.template; deep-merge into user's settings.json (UNION permissions.allow + enabledPlugins + extraKnownMarketplaces; APPEND hooks.PreToolUse; keep existing statusLine / autoUpdatesChannel / skipDangerousModePermissionPrompt). 11. CONFIG.JSON — load .claude/config.json.template; merge mcpServers by key WITHOUT overwriting existing servers. Remind me to fill , , . 12. INSTALL PLUGINS — run /plugin install for each of: superpowers@claude-plugins-official document-skills@ando-marketplace frontend-design@claude-plugins-official n8n-mcp-skills@n8n-mcp-skills apify-ultimate-scraper@apify-agent-skills rust-analyzer-lsp@claude-plugins-official hugging-face-cli@huggingface-skills ui-ux-pro-max@ui-ux-pro-max-skill marketing-skills@marketingskills codex@openai-codex claude-mem@thedotmack OPTIONAL (ask first, skip if marketplace unknown): browser-use, career-ops, SEO — install if I have their marketplaces registered; otherwise note as a TODO in the final table. 13. VERIFY — count skills (>=17), agents (>=6), commands (>=29); grep Caveman Mode / RTK / graphify in CLAUDE.md; list enabled plugins. 14. FINAL REPORT — table of v/!/x per section + reminder to fill MCP creds, review CLAUDE.md, reopen terminal, run /plugin. 15. CONFIGURATION GUIDE — after setting up everything, give me a concise guide for configuring anything in this bundle on MY machine. Reference real paths written to (~/.claude/CLAUDE.md, rules/, skills//SKILL.md, agents/.md, commands//.md, settings.json, config.json) and cover: - Add / remove / rename skills, agents, commands. - Enable / disable installed plugins and register new marketplaces. - Edit CLAUDE.md and rules safely (Caveman intensity, RTK, graphify pointer, TechRex Bundle markers). - Fill MCP credentials (search-console OAuth, n8n-mcp N8N_API_URL / N8N_API_KEY / dist path). - Add a new MCP server to config.json. - Tweak or disable the PreToolUse graphify hook. - Enable / disable / swap the statusline. - Restore the pre-install ~/.claude from the backup if I want to roll back. One page, code blocks where helpful, plain language. Do not invent steps. Do not run destructive commands. Ask before installing the statusline. If any step fails, stop and report — do not continue blindly. ``` --- ## Part 2 — Plugins installed (11) The agent installs these via `/plugin install `. Each lives on a public Claude Code marketplace. | Slug | Marketplace | Purpose | |------|-------------|---------| | `superpowers` | `claude-plugins-official` | TDD, debugging, process-discipline skills | | `document-skills` | `ando-marketplace` | PDF / DOCX / XLSX skills | | `frontend-design` | `claude-plugins-official` | Design-system helpers | | `n8n-mcp-skills` | `n8n-mcp-skills` | n8n workflow skills | | `apify-ultimate-scraper` | `apify-agent-skills` | Web scraping | | `rust-analyzer-lsp` | `claude-plugins-official` | Rust LSP integration | | `hugging-face-cli` | `huggingface-skills` | HF model / dataset ops | | `ui-ux-pro-max` | `ui-ux-pro-max-skill` | UI/UX design skills | | `marketing-skills` | `marketingskills` | Marketing + SEO content | | `codex` | `openai-codex` | OpenAI Codex integration | | `claude-mem` | `thedotmack` | Persistent memory | **Optional plugins** (not auto-installed — add manually if you have their marketplace registered): `browser-use`, `career-ops`, standalone `seo` plugin. The agent prompt surfaces these as TODOs in the final table so you can install them yourself with `/plugin install @`. Two extra marketplaces are auto-registered by the settings template so plugin installs resolve without manual marketplace adds: - `openai-codex` -> `github:openai/codex-plugin-cc` - `thedotmack` -> `github:thedotmack/claude-mem` --- ## Part 3 — Custom skills (17) Copied from the repo's `.claude/skills/` into `~/.claude/skills/`: `autoresearch` · `caveman` · `data-analyst` · `data-storytelling` · `design-system` · `exploratory-data-analysis` · `gmaps-scraper` · `graphify` · `kpi-dashboard-design` · `linkedin-profile-search` · `prompt-refine` · `seedance` · `senior-data-scientist` · `skill-development` · `social-media` · `sql-optimization-patterns` · `youtube-content-generator` > Note: The CCG quality-gate skill has been removed from the bundle. > The `superpowers` plugin now covers TDD / debugging / verification > territory more comprehensively. --- ## Part 4 — Sub-agents (6) Copied to `~/.claude/agents/`: `blog-writer-expert` · `comprehensive-qa-tester` · `fullstack-expert` · `system-troubleshooter` · `ui-ux-design-expert` · `youtube-content-strategist` --- ## Part 5 — Slash commands (29) Copied to `~/.claude/commands/`: - **`gsd/*` (26)** — new-project, create-roadmap, execute-phase, progress, status, check-todos, plan-fix, verify-work, ... - **`vibe-*` (3)** — vibe-build, vibe-coding, vibe-research. --- ## Part 6 — PreToolUse hook Auto-injected into `settings.json` under `hooks.PreToolUse`. Reminds Claude to read the graphify knowledge graph (if one exists) before doing raw Glob / Grep searches: ```json { "matcher": "Glob|Grep", "hooks": [{ "type": "command", "command": "[ -f graphify-out/graph.json ] && echo '{\"hookSpecificOutput\":{\"hookEventName\":\"PreToolUse\",\"additionalContext\":\"graphify: Knowledge graph exists. Read graphify-out/GRAPH_REPORT.md for god nodes and community structure before searching raw files.\"}}' || true" }] } ``` --- ## Part 7 — MCP server stubs Merged into `config.json` without overwriting your existing MCP servers. Replace the placeholders after install: ```json { "mcpServers": { "search-console": { "command": "npx", "args": ["-y", "search-console-mcp"] }, "n8n-mcp": { "command": "node", "args": [""], "env": { "MCP_MODE": "stdio", "LOG_LEVEL": "error", "DISABLE_CONSOLE_OUTPUT": "true", "N8N_API_URL": "", "N8N_API_KEY": "" } } } } ``` - **search-console** prompts for Google OAuth on first use. No code changes required. - **n8n-mcp** needs a running n8n instance (local, docker, or exposed via ngrok). Clone https://github.com/czlonkowski/n8n-mcp, build it, then point `args[0]` at the built entry file. --- ## Part 8 — Verification checklist After the agent finishes, open a fresh terminal and run: ```bash # Linux / macOS / Git Bash ls ~/.claude/skills | wc -l # >= 18 find ~/.claude/agents -name '*.md' | wc -l # >= 10 find ~/.claude/commands -name '*.md' | wc -l # >= 57 grep -c "Caveman Mode" ~/.claude/CLAUDE.md # >= 1 grep -c "RTK" ~/.claude/CLAUDE.md # >= 1 grep -c "graphify" ~/.claude/CLAUDE.md # >= 1 ``` Then start `claude` and run `/plugin` — you should see 10 enabled plugins. --- ## Part 9 — Removing the bundle - **Skills / agents / commands** — delete the copied files/dirs. - **CLAUDE.md** — delete everything between `=== BEGIN/END TechRex Bundle ===`. - **Plugins** — `/plugin uninstall ` for each. - **MCP servers** — drop the `search-console` and `n8n-mcp` entries from `config.json`. --- Curated by [TechRex](https://www.youtube.com/@The_TechRex) — AI coding agent tutorials for developers. Source: MIT-licensed on [GitHub](https://github.com/saiteja007-mv/techrex-claude-setup).