Tooling for AI-assisted development — behavioral guidelines, slash commands, and repo governance. Primarily built for Claude Code; guidelines are compatible with Cursor, GitHub Copilot, and Gemini CLI via AGENTS.md.
.claude/commands/ Slash commands — available as /command-name in any Claude Code project after setup
tools/ Setup and sync scripts
configs/ Config templates for AI coding assistants
notes/ Research notes, evaluations, tool comparisons
- AGENTS.md — AI agent behavioral guidelines. Read by Claude Code, Cursor, Copilot, and Gemini CLI.
- global-claude.md — Canonical behavioral rules imported by all Claude Code projects.
No git clone required:
curl -fsSL https://raw.githubusercontent.com/longieirl/ai-tools/main/tools/setup-claude.sh | bashInstalls to ~/.claude/:
global-claude.md— behavioral guidelines, loaded by all Claude Code projectscommands/— all slash commands available in every Claude Code session
curl -fsSL https://raw.githubusercontent.com/longieirl/ai-tools/main/tools/sync-project.sh | bashCreates AGENTS.md and CLAUDE.md in the current directory, wired to the upstream source.
Add to the top of the project's CLAUDE.md:
@~/.claude/global-claude.md
curl -fsSL https://raw.githubusercontent.com/longieirl/ai-tools/main/tools/setup-claude.sh | bashOr from within any Claude Code project: /sync-ai-config.
Slash commands available in any Claude Code session after running setup-claude.sh.
| Command | File | Description |
|---|---|---|
/update-all |
update-all.md | Update everything — global commands, behavioral guidelines, marketplace plugins, and local-install skills |
/sync-ai-config |
sync-ai-config.md | Sync this project's AI config files from the upstream repo |
/update-global-config |
update-global-config.md | Update ~/.claude/global-claude.md with latest behavioral guidelines |
/github-repo-lockdown |
github-repo-lockdown.md | Lock down a public GitHub repo — rulesets, CODEOWNERS, CI validation, security hardening |
/setup-dead-weight-audit |
setup-dead-weight-audit.md | Audit AI setup files for dead-weight instructions that produce no observable difference |
/security-audit |
security-audit.md | OWASP Top 10 application security audit — secrets, injection, auth, CVEs |
/dns-audit |
dns-audit.md | DNS and email security audit — DKIM, DMARC, SPF, CAA, MTA-STS across hosting panels |
All commands live in .claude/commands/. setup-claude.sh downloads them to ~/.claude/commands/ so they're available globally. To add a command: create a .md file in .claude/commands/, add it to the loop in tools/setup-claude.sh, re-run setup.
Skills come in three types with different update paths:
- Built-in — shipped with Claude Code or Superpowers; updated automatically
- Marketplace plugin — declared in
enabledPlugins; update via/plugin updatein any Claude Code session - Local install — cloned or npx-installed into
.claude/skills/; update by re-running the install command
If you copy-paste a skill's .md file directly into your project, you own that file. It becomes a fork — no auto-update. Track the upstream source URL in the skill's frontmatter and re-paste manually when you want changes.
| File | Description | Type | Update |
|---|---|---|---|
| impeccable | Production-grade frontend interface design — craft, audit, animate, polish | Local install | npx impeccable@latest install |
| taste-skill | Metric-based UI/UX engineering — React/Next.js, Tailwind, Framer Motion, no AI-slop | Local install | Re-clone from source |
| emil-design-eng | Design engineering review — Before/After/Why format, animation framework, CSS mastery | Local install | Re-clone from source |
| ui-ux-pro-max | Design intelligence database — 67 styles, 96 palettes, 57 font pairings, 25 chart types | Marketplace plugin | /plugin update |
| update-config | Configure Claude Code settings.json — permissions, hooks, env vars, MCP servers | Built-in | Auto (with Claude Code) |
| github:secure-repo | Guided GitHub repo security hardening — rulesets, CODEOWNERS, access controls | Marketplace plugin | /plugin update |
| gsd:progress | GSD Redux — auto-detect and run next step in the plan→execute→verify→ship loop | Local install | npx @opengsd/get-shit-done-redux@latest |
| static-analysis:semgrep | Semgrep static analysis — security vulnerabilities, OWASP, custom rules | Marketplace plugin | /plugin update |
| ralph-wiggum | Iterative AI development loops — same prompt fed repeatedly so Claude builds on its own previous work | Marketplace plugin (claude-code-plugins) |
/plugin update |
| File | Description |
|---|---|
| shadcn | shadcn/ui MCP — install components, browse registry, query docs |
| github | GitHub API MCP — PRs, issues, file read/write, repo search |
| context7 | Live library documentation — current API docs and code examples for any package |
| ide-diagnostics | IDE diagnostics MCP — TypeScript errors, lint warnings via language server |
| File | Description |
|---|---|
| rtk-token-saving.md | RTK (Rust Token Killer) — CLI proxy that cuts shell-command tokens 60–90% before they hit LLM context |
| dns-email-security.md | DNS email security records — DKIM, DMARC, CAA, MTA-STS, TLS-RPT — why each matters and how to implement them |
| github-repo-standards.md | Non-negotiable GitHub repo standards — PR auto-assign, branch auto-delete, lean workflow output |
| gsd | Get Shit Done Redux — spec-driven dev system that solves context rot with parallel subagent execution |
| playwright | Browser automation — E2E tests, accessibility, cross-browser, responsive, screenshots |
| lighthouse | Automated auditing — performance, accessibility, SEO, best practices |
| gitleaks | Secret scanning — detects hardcoded credentials in git history and working tree |
| OpenSpec | Spec-driven feature planning — generates proposal docs, design decisions, and tasks that persist alongside code across sessions and agents |
After cloning, run setup:
bash tools/setup-claude.sh
git config core.hooksPath .github/hookssetup-claude.sh downloads guidelines and commands into ~/.claude/. core.hooksPath enforces local protection on main (no direct commits or pushes).