By ggrace519
Claude Code Dev Studio core: 5 generalist agents (plan, review, security, test, deploy) plus the cross-cutting skills every project uses.
Deployment readiness and pre-production checklist specialist. Use proactively before any production deployment, environment promotion, release cut, or go/no-go assessment.
Universal system design and architecture planning specialist. Use proactively when defining component boundaries, data flows, integration patterns, or making major structural decisions before or during implementation.
Pull request and code review specialist. Use proactively when a PR is opened, a diff is ready for review, or the user requests feedback on a changeset before merging.
Security audit and hardening specialist. Use proactively at the start of the Hardening phase and whenever security-sensitive code is written — authentication, authorization, cryptography, file I/O, secrets handling, or processing untrusted external data.
Test writing and execution specialist. Use proactively after implementation is complete, after a PR review resolves issues, or when test coverage for a module or feature needs improving.
API design and implementation specialist. Auto-invoked when writing HTTP endpoints, REST or GraphQL APIs, authentication flows, API clients, webhooks, or defining data contracts between services.
Code review reference — the seven review dimensions and comment severity labels. Use proactively when reviewing a diff or self-reviewing changes before handing off to the pr-code-reviewer agent.
Accessibility (WCAG 2.2 AA+, ARIA, assistive-tech compatibility, keyboard / switch / voice / screen-reader UX). Auto-invoked when building UI, prepping for audit, or remediating a11y findings.
Internationalization and localization — string externalization, ICU MessageFormat, plurals, RTL, locale-aware formatting, translation workflow. Auto-invoked when adding locales, debugging localized bugs, or designing the i18n architecture.
Cross-channel user notifications — push, email, SMS, in-app, webhook, batching / throttling / quiet-hours, preferences, deliverability. Auto-invoked when designing notification flows, investigating open-rate drops, or implementing preferences UI.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A universal, stack-agnostic Claude Code playbook: 19 always-on agents (14 domain agents + 5 core generalists) plus a library of ~90 skills the agents compose on demand. Aligned with NIST SSDF (SP 800-218) and phase-gated Agile delivery. See DECISIONS.md ADR-0007 for the architecture.
| File | Purpose |
|---|---|
CLAUDE.md | The playbook — phases, the agent/skill model, prefix registry |
.claude/agents/ | The 19 always-on agent definitions (14 domain + 5 core) |
skills/ | The skill library (<name>/SKILL.md) — domain + cross-cutting |
catalog.json | Index of agents and skills (name, pack, kind, scope, model, description) for JIT selection |
bin/ccds.{ps1,sh} | Dispatcher — sync, verify, update, uninstall, version |
Install-Playbook.ps1 | Windows installer (stage/promote, SHA256-verified, PATH-aware, CLAUDE.md block) |
install-playbook.sh | Linux/macOS installer (same behavior, shell-rc PATH block) |
Sync-AgentPacks.{ps1,sh} | Stage a pack's domain skills into a project (invoked by ccds sync) |
Verify-Agents.ps1 / verify-agents.sh | Validate agents and skills against ADR-0001 invariants |
build-release.ps1 | Build reproducible release ZIP + sidecar SHA256 |
scripts/jit-claude.md | Canonical source for the ccds pointer block injected into ~/.claude/CLAUDE.md |
.github/workflows/release.yml | Tag-driven release build + GitHub Release publication |
DECISIONS.md | Architecture decision records (ADR-0001 … ADR-0009) |
CHANGELOG.md | Session-by-session history |
CONTRIBUTING.md | Contribution terms |
LICENSE | PolyForm Noncommercial 1.0.0 |
~/.claude/
agents/ ← 19 always-on agents (14 domain + 5 core; always loaded)
skills/ ← cross-cutting skills (always available)
playbook/ ← managed by installer
bin/ ← ccds dispatcher (ccds.ps1 / ccds.sh / ccds symlink)
scripts/ ← Sync-AgentPacks, Verify-Agents, jit-claude.md
agents/ ← source copy of the 19 agents
skills/ ← full skill library (domain skills copied to projects on demand)
catalog.json ← agent + skill index for JIT selection
version.txt
README.md
CLAUDE.md ← your global Claude instructions; installer injects the ccds pointer block
The 19 agents are cheap enough (~850 tokens of trimmed descriptions) to load every session, so there is no agent-activation step. Each domain agent composes its <pack>-* skills via the Skill tool — and because subagents can invoke skills but cannot spawn other subagents, one domain agent handles a multi-specialty task in a single coherent context.
Skills are the just-in-time layer. The sync-agents skill (or ccds sync):
~/.claude/playbook/catalog.json — agents + skills, lightweight metadata only./.claude/skills/ in the projectCross-cutting skills (playbook-conventions, api-design, ux-design, security-checklist, code-review-checklist, common-*) install once to ~/.claude/skills/ and are always available.
The repo doubles as a Claude Code plugin marketplace: one plugin per pack
(ccds-saas, ccds-ai, …) plus ccds-core (the 5 core agents + cross-cutting
skills). Claude Code handles versioning, updates, and enable/disable — no
installer, no PATH, no restart dance.
/plugin marketplace add ggrace519/claude-code-dev-studio
/plugin install ccds-core@ccds
/plugin install ccds-saas@ccds # one per archetype you work in
Update later with /plugin marketplace update ccds. The marketplace tree
(.claude-plugin/marketplace.json + plugins/) is generated from the library
source by scripts/build-marketplace.py and gated for freshness in CI.
The ZIP installer below remains fully supported — it additionally provides the
ccds CLI, the global ~/.claude/playbook/ library, and per-project skill
staging via ccds sync.
The installer downloads a GitHub Release ZIP, verifies its SHA256 against the sidecar, stages to <prefix>.new, snapshots the existing install to <prefix>.previous, and atomically promotes. It copies the 19 agents to ~/.claude/agents/, the cross-cutting skills to ~/.claude/skills/, injects the ccds block into ~/.claude/CLAUDE.md, and updates PATH so ccds resolves in new shells.
Windows (PowerShell 5.1 or 7+):
iwr -UseBasicParsing https://raw.githubusercontent.com/ggrace519/claude-code-dev-studio/main/Install-Playbook.ps1 | iex
npx claudepluginhub ggrace519/claude-code-dev-studio --plugin ccds-coreAI / LLM application domain specialist. Ships the ai-architect agent and 6 ai-* skills it composes.
Media / streaming domain specialist. Ships the media-architect agent and 6 media-* skills it composes.
Data Platform domain specialist. Ships the dataplat-architect agent and 7 dataplat-* skills it composes.
DevTool / CLI / Library domain specialist. Ships the devtool-architect agent and 5 devtool-* skills it composes.
Desktop App domain specialist. Ships the desktop-architect agent and 5 desktop-* skills it composes.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Develop, test, build, and deploy Godot 4.x games with Claude Code. Includes GdUnit4 testing, web/desktop exports, CI/CD pipelines, and deployment to Vercel/GitHub Pages/itch.io.
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Comprehensive PR review agents specializing in comments, tests, error handling, type design, code quality, and code simplification
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.