Open source · MIT· formerly Claude Arsenal

One spellbook,
every runtime.

95 cross-runtime skills and 7 specialist agents for Claude Code, Codex, and multi-agent workflows — validated, cataloged, and installed with a single incantation.

95 SKILLS ✦ 7 AGENTS ✦ ONE COMMAND ✦ CROSS-RUNTIME
install.sh — one-line install
$
curl -fsSL https://raw.githubusercontent.com/majiayu000/spellbook/main/install.sh | bash -s -- --target all
# installs into ~/.claude/skills (Claude Code) and ~/.agents/skills (Codex)
95
Cross-runtime skills
7
Specialist agents
2
Runtimes, one source tree
MIT
Use freely
systematic-debugging ✦ vibeguard ✦ multi-model-orchestrator ✦ app-ui-design ✦ typescript-project ✦ release-engineering ✦ github-trending ✦ codebase-audit ✦ threads ✦ prd-master ✦ figma-to-code ✦ xiaohongshu ✦ rust-project ✦ flowguard ✦ api-design ✦ humanizer ✦systematic-debugging ✦ vibeguard ✦ multi-model-orchestrator ✦ app-ui-design ✦ typescript-project ✦ release-engineering ✦ github-trending ✦ codebase-audit ✦ threads ✦ prd-master ✦ figma-to-code ✦ xiaohongshu ✦ rust-project ✦ flowguard ✦ api-design ✦ humanizer ✦
product-discovery ✦ test-driven-development ✦ golang-web ✦ review-gate ✦ slides ✦ gpu-use ✦ auth-security ✦ skill-creator ✦ harmonyos-app ✦ multi-ai-research ✦ devops-excellence ✦ git-commit-smart ✦ clash-doctor ✦ strategic-compact ✦ brainstorming ✦ database-patterns ✦product-discovery ✦ test-driven-development ✦ golang-web ✦ review-gate ✦ slides ✦ gpu-use ✦ auth-security ✦ skill-creator ✦ harmonyos-app ✦ multi-ai-research ✦ devops-excellence ✦ git-commit-smart ✦ clash-doctor ✦ strategic-compact ✦ brainstorming ✦ database-patterns ✦
01/ Why Spellbook

Write once, cast anywhere.

Cross-runtime install

One source tree installs into both Claude Code and Codex. Write a skill once, use it from any coding agent.

Validated registry

Every installable skill is checked by the registry validator before it ships — structure, frontmatter, dependencies.

Progressive disclosure

Larger skills ship references, templates, scripts, and evals — not one giant prompt that eats your context.

Practical coverage

Engineering, operations, product, UI, content, and agent workflows — one catalog, battle-tested in production.

02/ The catalog

Twelve schools of magic.

full skill registry ↗

Every skill is a directory with hard rules, practical examples, and verification checklists. Pick a school, or install them all.

Tooling & Automation

15 skills

Development Architecture

7 skills

Development Practices

5 skills

Operations & Deploy

5 skills
$ python3 scripts/validate_skills.py search --tag react --jsonThe registry is machine-searchable — free text, tags, language filters.
03/ Pick a workflow

Start small. Bundle by job.

Install a job-shaped set of skills, then add more once the workflow sticks.

Frontend & UI

Product UI, landing pages, design systems, Figma handoff.

./install.sh --target all --skills frontend-design,app-ui-design,ui-design-system,figma-to-react

Code quality

Audits, bug fixes, refactors, root-cause debugging.

./install.sh --target all --skills codebase-audit,flowguard,systematic-debugging,review-gate

Ops & deploy

Shipping apps, hardening servers, diagnosing local and network issues.

./install.sh --target all --skills release-engineering,server-security,clash-doctor,system-doctor

Product & docs

Discovery, PRDs, technical specs, metrics plans.

./install.sh --target all --skills product-discovery,prd-master,technical-spec,product-analytics

Agent workflows

Cross-review, multi-AI research, context handoff, anti-hallucination checks.

./install.sh --target all --skills codex-agent,multi-ai-research,strategic-compact,vibeguard
04/ Agents

Seven familiars.

Specialist Claude Code agents for coordination, exploration, review, security, and infrastructure.

05/ The agent infra stack

Part of a larger circle.

Spellbook is the Extend layer of an open-source stack for running coding agents as serious infrastructure. Every piece works standalone; together they close the loop.