**CLAUDE CODE DEVELOPER — AI Content Production Pipeline (LinkedIn Agency)**

**Overview**

We're a LinkedIn content and lead generation agency working with 47 clients. Our entire operation runs on Claude AI — we've built a suite of 40+ custom skills (structured markdown prompt files) that handle everything from content strategy to visual production briefs to carousel generation.

The problem: our production workflow involves too many manual steps. A tea ---------- mber runs each skill sequentially, copies outputs between tools, creates folders manually, and pushes briefs into Canva by hand. At 47 clients producing multiple content assets per month each, that's a lot of admin that shouldn't require a human.

We want to build a Claude Code pipeline that chains these steps automatically — so our designer focuses on QA and finishing, not copy-pasting and folder management.

---

**Current manual workflow**

Here's what happens right now, step by step:

1. Approved written content is taken from the client project
2. Tea ---------- mber opens Claude, loads the Visual Architect skill, inputs the content ? receives structured Nano Banana image prompts (formatted for Manus image generation)
3. Tea ---------- mber opens Claude, loads the Carousel Visualiser skill, inputs the content ? receives a carousel brief with slide-by-slide copy and layout direction
4. Carousel brief is manually pushed into Canva via the Canva MCP connector
5. Nano Banana prompts are manually taken into Manus for image generation
6. Finished assets are manually dropped into the correct Google Drive client folder

We want steps 2, 3, 4, and 6 automated. Step 5 (Manus image generation) remains manual for now.

---

**What we want to build**

A Claude Code script (or lightweight CLI tool) with a simple trigger. The designer inputs:

- Client name
- Approved written content (pasted or file path)

And the pipeline automatically:

1. **Runs the Visual Architect skill** against the content ? produces formatted Nano Banana prompts ready for Manus, saved to a structured output file
2. **Runs the Carousel Visualiser skill** against the content ? produces carousel brief with slide copy, layout direction, and content QA output
3. **Pushes the carousel brief into Canva** via Canva MCP — creating the design file in the correct client workspace
4. **Locates or creates the correct Google Drive client folder** and drops all outputs (prompts file, carousel brief, Canva link) into it, correctly named

The designer then opens the folder, takes the Nano Banana prompts into Manus manually, and does final QA on the Canva output.

---

**The tech stack**

- **Claude Code** — primary development environment
- **Anthropic API** — claude-sonnet ---------- , skills loaded as system prompts
- **Canva MCP** — already connected ( ---------- /mcp), used for pushing carousel briefs and creating design files
- **Google Drive MCP** — already connected ( ---------- — Drive access enabled), used for folder management and file storage
- **Our skill files** — 40+ structured .md files, two of which are central to this pipeline (Visual Architect v4.0, Carousel Visualiser v2.1). You'll receive full access
- **Google Drive** — client folders already exist and follow a consistent naming convention

---

**Specific deliverables**

1. A working Claude Code pipeline script that executes the full automated sequence described above
2. A simple interface for the designer — ideally a CLI prompt or basic input form. No complex UI required
3. Error handling — if a client folder doesn't exist in Drive, create it. If Canva push fails, log it and flag to the designer
4. Output logging — a simple run log per execution (client name, date, assets created, Canva link, Drive folder path)
5. Documentation — a plain-English setup and usage guide written for a non-technical designer. Step by step, no assumptions
6. A short walkthrough video of the working pipeline (Loom is fine)

---

**What we're not building**

- A full web app or dashboard
- Anything that touches Manus (that stays manual for now)
- Anything client-facing
- Complex UI — this is an internal production tool

---

**Scale context**

47 clients. Each client produces approximately:
- 10 personal LinkedIn posts per month
- 4 carousels per month
- 8 company page posts per month

The pipeline needs to handle single-client runs and ideally batch runs (multiple clients queued). Not all of this in v1 — but worth knowing so you don't architect something that can't scale.

---

**Who we're looking for**

- Solid experience with Claude Code and the Anthropic API
- Has worked with MCP integrations before — ideally Canva or Google Drive
- Comfortable reading and working with structured markdown prompt files (our skills)
- Builds things that non-technical people can actually use
- Asks good questions before building, not after
- Available for a short discovery call before starting

---

**To apply**

Please answer these three questions in your proposal:

1. Have you built anything with Claude Code before? Describe the project and what it did.
2. Have you worked with MCP integrations — specifically Canva MCP or Google Drive MCP? What was the use case?
3. How would you approach chaining multiple large system prompts (each 1,000–3,000 words) through the Anthropic API without hitting context issues?

Proposals without answers to all three will not be considered.
Let me know how much you think the project would cost