# create-claudify Scaffold a complete [Claudify Content Specialist](https://claudify.tech) workspace in one command. ```bash npx create-claudify ``` Claudify Content Specialist is a Claude Code operating system for content production, brand voice, and audience growth. It ships as a self-contained workspace with agents, hooks, commands, and memory — ready to produce content across X, LinkedIn, Instagram, Threads, newsletters, blog posts, and more. --- ## What gets created Running `npx create-claudify` scaffolds a full workspace: ``` my-content-os/ ├── CLAUDE.md Context file — Claude reads this every session ├── CLAUDE.local.md Personal overrides (gitignored) ├── Task Board.md Current work tracking ├── .gitignore ├── .claude/ │ ├── memory.md Session state (6-tier memory architecture) │ ├── knowledge-base.md Content rules (30+ pre-seeded hard rules) │ ├── knowledge-nominations.md Candidate learnings pipeline │ ├── command-index.md All commands with triggers and tools │ ├── settings.json 14 hooks wired to Claude Code lifecycle │ ├── agents/ 10 specialist subagents │ │ ├── auditor.md │ │ ├── content-creator.md │ │ ├── brand-voice-keeper.md │ │ ├── content-strategist.md │ │ ├── newsletter-orchestrator.md │ │ ├── drafter.md │ │ ├── prompt-architect.md │ │ ├── newsletter-auditor.md │ │ ├── stylist.md │ │ └── analyst.md │ ├── commands/ 13 slash commands │ │ ├── start.md │ │ ├── sync.md │ │ ├── wrap-up.md │ │ ├── safe-clear.md │ │ ├── feedback.md │ │ ├── content.md │ │ ├── post.md │ │ ├── thread.md │ │ ├── blog.md │ │ ├── newsletter.md │ │ ├── launch-post.md │ │ ├── readme.md │ │ └── audit.md │ ├── hooks/ 14 hooks (9 base + 5 newsletter) │ │ ├── guard-bash.sh │ │ ├── backup-before-write.sh │ │ ├── completeness-gate.sh │ │ ├── log-changes.sh │ │ ├── log-failures.sh │ │ ├── log-stop-verdict.sh │ │ ├── pre-compact-handoff.sh │ │ ├── post-compact-resume.sh │ │ ├── session-reset.sh │ │ ├── newsletter-em-dash-check.sh │ │ ├── newsletter-jargon-check.sh │ │ ├── newsletter-schema-check.sh │ │ ├── newsletter-contrast-check.sh │ │ └── newsletter-paragraph-collapse-check.sh │ └── agent-memory/ │ ├── content-creator/ │ │ ├── MEMORY.md │ │ └── brand-memories/ │ │ ├── _TEMPLATE.md │ │ └── {your-brand}.md (one per brand you entered) │ └── newsletter-orchestrator/ │ └── MEMORY.md └── newsletter/ (if you chose to include it) ├── README.md Start here for the newsletter system ├── SYSTEM.md Full operational guide ├── ISSUE-SCHEMA.md ├── GOD-PROMPT-RUBRIC.md ├── BACKLOG.md ├── COMPONENTS.html ├── email-template.html CUSTOMIZE this before your first issue ├── build.py ├── send.js ├── render-test.sh ├── test-god-prompt.py ├── .env.example Copy to .env and fill in your keys ├── foundations/ │ ├── PERSONA.md CUSTOMIZE for your audience │ ├── VOICE.md CUSTOMIZE for your voice │ ├── GLOSSARY.md CUSTOMIZE for your terms │ ├── BUDGETS.md │ ├── PIPELINE-PERSONALITIES.md │ └── MISTAKES-LOG.md ├── issues/ ├── briefs/ └── drafts/ ``` --- ## Interactive setup The CLI asks 6 questions: 1. **Workspace directory** — where to create the workspace (default: `./my-content-os`) 2. **Brand names** — one or more brands you produce content for (e.g. `Acme`, `acme-blog`) 3. **Newsletter** — whether to include the full newsletter production system 4. **Platforms** — which platforms you publish to (X, LinkedIn, Instagram, Threads, TikTok, YouTube, Beehiiv, blog, Product Hunt, Hacker News) 5. **Default tone** — direct / conversational / authoritative 6. **Long-form length** — short (~600w) / medium (~1200w) / long (~2000w) --- ## Quick start After scaffolding: ```bash cd my-content-os code . # or your editor of choice ``` Open Claude Code in the workspace. Then: ``` /start — begin your session (reads memory, sets context) /post X "your topic" — draft a single X post /thread "your topic" — draft an X thread /blog "keyword" — draft a blog post (SEO-aware) /content mybrand — run the full content pipeline /newsletter strategy 001 — start your first newsletter issue ``` --- ## System components ### 10 specialist agents | Agent | Responsibility | |---|---| | `content-creator` | Lead production agent across all 7 content modes and 9 platforms | | `content-strategist` | Topic selection, content calendar, brief production | | `brand-voice-keeper` | Voice consistency across sessions, drift detection | | `auditor` | Quality gate — reviews all output in fresh context | | `newsletter-orchestrator` | Coordinates the 9-phase newsletter pipeline | | `drafter` | Writes newsletter issue body content | | `prompt-architect` | Designs god-tier prompts, tests against 3 codebases | | `newsletter-auditor` | 25 mechanical + 13 semantic + regression checks | | `stylist` | Visual rendering gate — 20 checks + Playwright at 4 viewports | | `analyst` | Post-send learning — day-of-send + day-7 reports | ### 14 hooks Hooks run automatically in Claude Code lifecycle events: | Hook | Trigger | What it does | |---|---|---| | `guard-bash` | PreToolUse Bash | Blocks dangerous commands (rm -rf /, force push, secret exposure) | | `backup-before-write` | PreToolUse Write | Creates timestamped backups, 7-day auto-pruning | | `completeness-gate` | PreToolUse Write\|Edit | Checks knowledge-base provenance, memory line limit, JSON validity | | `log-changes` | PostToolUse Write\|Edit | Audit trail for all file changes | | `log-failures` | PostToolUseFailure | Categorised failure logging | | `log-stop-verdict` | Stop | JSONL verdict log, quality gate activation | | `pre-compact-handoff` | PreCompact | Saves state before context compaction | | `post-compact-resume` | SessionStart(compact) | Restores context after compaction | | `session-reset` | SessionStart(user) | Resets gate files, validates agent frontmatter | | `newsletter-em-dash-check` | PreToolUse Write | Blocks em dash in newsletter issues | | `newsletter-jargon-check` | PreToolUse Write | Blocks banned acronyms and filler phrases | | `newsletter-schema-check` | PreToolUse Write | Validates required frontmatter and section headers | | `newsletter-contrast-check` | PreToolUse Write | Blocks low-contrast hex values | | `newsletter-paragraph-collapse-check` | PreToolUse Write | Blocks `\n\n` inside `

` tags | ### 6-tier memory architecture 1. `memory.md` — active session context 2. Content Creator Memory — production log, calibration, approved patterns 3. Brand Memories — per-brand voice samples, ICP, signature phrases, banned phrases 4. Knowledge Base — system-wide content rules (auditor-gated) 5. Knowledge Nominations — candidate learnings pipeline 6. Daily Notes — chronological session history --- ## Newsletter system If you chose to include the newsletter system, read `newsletter/README.md` first. The newsletter system ships with: - **6 foundation files** (PERSONA, VOICE, GLOSSARY, BUDGETS, PIPELINE-PERSONALITIES, MISTAKES-LOG) - **15 pre-seeded documented failures** in MISTAKES-LOG.md with auditor checks - **Full build pipeline**: `build.py` → `render-test.sh` → `send.js` - **God-tier prompt validation**: 6-criterion rubric + 3-codebase test harness - **9-phase pipeline** coordinated by the `newsletter-orchestrator` agent **Customise the foundations before your first issue** — especially `PERSONA.md`, `VOICE.md`, and `email-template.html`. --- ## Requirements - [Claude Code](https://docs.anthropic.com/en/docs/claude-code) installed and authenticated - Node.js 18+ (for `npx create-claudify` and `send.js`) - Python 3.8+ (for `build.py` and `test-god-prompt.py`) - Playwright (for `render-test.sh` — installed on first run via `npx playwright`) --- ## License MIT