Prompt file fetched from GitHub and filled with your values

What it is

This n8n workflow keeps prompts in a GitHub repo and fills them in at run time. A manual trigger starts it, and a Set node holds the repo, the file’s folder and name, and the values for the prompt, such as company, product, features and sector. The GitHub node fetches the file and Extract From File reads it as text. A check lists any placeholder with no value, a Code node fills in the rest, and an AI agent on Ollama runs the result. Thirteen nodes do the work, beside four sticky notes.

Who it suits

Marketers and prompt writers who keep prompts as Markdown files in a repo and want to reuse each one with new values. Skip it if you have no prompt repo yet. As imported, the file name SEO/keyword_research.md points to a file you must write yourself.

You need a little n8n: import a file, add a GitHub credential and an Ollama credential, and write one prompt file with placeholders in it.

Where to use it

Any n8n that can reach GitHub and your Ollama server will run it. A private repo works when the credential can read it.

• n8n Cloud: import the file into a new workflow there, with an Ollama address it can reach
• Your own n8n server: the same file and steps, with Ollama on the same machine
• GitHub: a repo that holds your prompt files, one file for each prompt
• Other triggers: replace the manual start with a form, a schedule or a webhook, and set the values from there

Setup

Download saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. A GitHub credential and an Ollama address are all it needs.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Add a GitHub credential and select it in GitHub.
3. Add an Ollama credential with your server’s base URL and select it in Ollama Chat Model. Pick the model you have pulled.
4. Open setVars. Replace <YOUR_GITHUB_ACCOUNT>, <YOUR_PROMPT_REPO> and <YOUR_COMPANY> with your own values. Set path to the folder of your prompt file, ending in a slash, and prompt to its name.
5. Write that prompt file in the repo, with placeholders such as {{ setVars.product }}. The word after the last dot must be a field of setVars.
6. Press Execute workflow. The model’s answer is in promptResponse, in Prompt Output.

How to use it

Put the values in setVars and press Execute workflow. The workflow fetches the file, finds every placeholder and checks that setVars has a field for each. If one is missing, the run stops with an error that names it. Otherwise the filled prompt goes to the agent.

Example

A repo holds SEO/keyword_research.md, a prompt that uses {{ setVars.product }} and {{ setVars.sector }}.

With product set to Manual Therapy and sector to physical therapy, the workflow swaps each placeholder for those words. AI Agent receives the finished prompt, and Prompt Output returns the model’s answer. A placeholder with no matching field, such as {{ setVars.budget }}, stops the run and names budget.

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. This one is plumbing around a single idea: the prompt lives in Git, not in n8n. Run it once with a placeholder missing on purpose to see the stop message.

For experts. Placeholders are matched by their last word, so {{ setVars.product }} and {{ product }} both read the product field.

Good to know

The original copy held its author’s account and repo, and the values it filled in did not match its own check.

• Repairs: the original copy named its author’s GitHub account, a repo and a business, which Mindket’s copy replaces with <YOUR_GITHUB_ACCOUNT>, <YOUR_PROMPT_REPO> and <YOUR_COMPANY>. Its replace step also had features and a keyword written into the code, so the values in setVars were checked but not used. Mindket’s copy reads every field from setVars
• Prompt file: SEO/keyword_research.md is a file of the original copy’s repo. Write your own file or point path and prompt at one you have
• Model: Ollama Chat Model sets none, so the node uses its default, llama3.2. Pull it or pick another
• Missing values: the run stops with an error that lists each placeholder with no field
• Fresh text: the file is fetched on every run, so an edit in the repo takes effect at once
• Your data: the prompt and your values go to your Ollama server, and GitHub sees only the file request
• Credentials: the JSON holds none