Stop writing every post from scratch. Configure your brand once. Then your AI agent runs the operation and you approve what ships:

  • Automate drafting: one brief becomes ready-to-review drafts for up to a week, across X, LinkedIn, Instagram, TikTok & YouTube, each platform-correct and in your voice.
  • Automate niche research: the agent finds the posts that actually outperformed in your niche and extracts their patterns.
  • Automate production planning: every approved video script gets a storyboard, ready-to-run asset prompts, and a staged checklist.
  • Keep control: your never-say rules are hard constraints, and nothing publishes without you.

One download with your free customer account, free for commercial use. Open it with your agent and ship up to a week of content from your first brief.

How to set it up

  1. Download and unzip. You get a folder called ai-social-media-starter: 21 plain-text files, no code, no installs, no API keys.
  2. Open it with your AI agent. The workspace is agent-agnostic: Claude Code, Codex, Cursor, or whatever agentic system you already use; any agent that can read files can run it. In Claude Code the included CLAUDE.md orients the agent automatically; in any other setup, provide the files and say "read CLAUDE.md first".
  3. Fill in your brand profile. Open context/brand-profile.md and complete it, or simply tell the agent: "Interview me against the brand profile template and fill it in from my answers." Already have brand documents (a brand guide, a positioning doc, website copy, past posts)? Drop them into the folder and tell the agent to distill the profile from them; it will confirm the result back to you. Don't skip the never-say list; it is the highest-leverage field in the file.
  4. Pick one or two platforms to start. Five platforms on day one is a recipe for mediocre everywhere. For founders and consultants, LinkedIn plus X is the natural opening pair.
  5. Run your first niche analysis. Workflow 01 walks you through it: pick 5–9 comparable accounts, find the posts that outperformed their own baseline, and let the agent extract the hook patterns, topic clusters, and avoid-rules. About an hour, and everything you generate afterwards stands on evidence instead of guesswork.
  6. Run your first content week. Write one brief (ten minutes of your thinking: the claim, the evidence, the platforms). The agent drafts the week: platform-correct, in your voice, grounded in your analysis. You review against the 8-point gate, publish what passes, and log it.
  7. For video pieces, follow the plan. A Reel, TikTok, or YouTube script that passes your review automatically gets a production plan: storyboard, asset prompts, action checklist. Film it yourself, or produce it with AI assistance; the plan covers both routes.

What this is

Most "AI content" advice hands you a pile of prompts. This resource hands you a working agent workspace, a folder you open with an AI agent (Claude Code, Codex, Cursor, or whatever agentic system you already use) that contains everything social media automation needs:

  • One brand profile every platform reads: your voice, audience, content pillars, and the never-say list that keeps AI output sounding like you instead of like everyone.
  • Four workflows: niche analysis (find the posts that actually outperformed in your niche and extract their patterns), the content week (one brief becomes a week of drafts), the review gate (an 8-point check every draft passes before you publish), and the script-to-production handoff.
  • Five platform specialists: dedicated agent definitions and playbooks for X, LinkedIn, Instagram, TikTok, and YouTube, each with the hook patterns, formats, signal metrics, scripting methods, and visual-prompt blueprints that are different on each platform.
  • A production planning system: when a video draft passes your review, the agent automatically writes a production plan covering a beat-by-beat storyboard, a dependency-ordered asset plan with ready-to-run image and video-scene prompts, and a staged checklist that tells both of you whose move it is.

Why agent workflows, not a prompt pack

Because the difference between generic AI output and posts that sound like you is context, not cleverer prompts. These agent workflows give your agent a memory (your brand profile, your content log, your niche analyses, your open production plans) and a contract: drafts only, your voice rules are hard constraints, you publish. That is the same architecture always-on content systems are built on: complete, yours, and ready to run this week.

What makes it different

  • The agent orients before acting. Every session starts with a workspace status: what exists, what's stale, which production plan is mid-flight and whose move it is. You don't need to remember the method; the agent carries it.
  • Evidence over intuition. Generation is grounded in your niche analysis and your own performance log, and the agent never invents engagement numbers.
  • Production-grade prompts. Image prompts are full direction (subject, setting, lighting, mood, palette) and video scene prompts choreograph the action second by second. No one-liner prompts, no AI-slop output.

It's free

Free under CC-BY-4.0: use it, adapt it, and ship it commercially. Attribution appreciated.