Prompt Action 🤖

A GitHub Action that lets you Prompt AI directly in your workflows.

• Prompt Action 🤖
  • Usage examples 🚀
    • Simple prompting
    • Structured Outputs
    • Templating with Variables
  • Permissions 🔒
  • Inputs ⚙️
  • Outputs 📤

Usage examples 🚀

Simple prompting

- uses: FidelusAleksander/prompt-action@v1
  with:
    prompt: 'What is the meaning of life?'

- uses: FidelusAleksander/prompt-action@v1
  with:
    system-prompt: 'You are Gilfoyle from Silicon Valley.'
    prompt: 'Tell me about your latest project.'

Structured Outputs

You can ensure the model returns data in a specific format by providing a JSON Schema.

- uses: FidelusAleksander/prompt-action@v1
  id: prompt
  with:
    prompt: |
      Will humanity reach Mars by 2035?
    response-schema-file: path/to/your-schema.json
- name: Use the output
  run: |
    echo "Response: ${{ fromJSON(steps.prompt.outputs.text).response }}"
    echo "Confidence: ${{ fromJSON(steps.prompt.outputs.text).confidence }}"
    echo "Tags: ${{ fromJSON(steps.prompt.outputs.text).tags }}"

Example schema

{
  "type": "object",
  "properties": {
    "response": {
      "type": "string",
      "description": "The main response text"
    },
    "confidence": {
      "type": "number",
      "minimum": 0,
      "maximum": 1,
      "description": "Confidence level from 0 to 1"
    },
    "tags": {
      "type": "array",
      "items": {
        "type": "string"
      },
      "description": "Relevant tags or categories"
    }
  },
  "required": ["response", "confidence", "tags"],
  "additionalProperties": false
}

Templating with Variables

You can create dynamic prompts using {{ variable }} syntax with the usage of the vars parameter.

- uses: FidelusAleksander/prompt-action@v1
  with:
    system-prompt: |
      You are a {{ language }} expert translator. 

      You will be provided with text to translate to {{ language }}.
      Respond with nothing but the translated text.

    prompt-file: README.md
    vars: |
      language: Spanish

For more advanced templating features like loops, conditionals, and filters, see the Nunjucks templating documentation.

Tip

Variable templating makes most sense when using prompt-file or system-prompt-file inputs, as it allows you to maintain reusable prompt templates with dynamic content.

Permissions 🔒

This actions requires at minimum the following permissions set.

permissions:
  contents: read
  models: read

Inputs ⚙️

Input	Description	Required	Default
prompt	Text that will be used as user prompt	No*	-
prompt-file	Path to a file containing the user prompt	No*	-
token	Personal access token	No	${{ github.token }}
model	The AI model to use. See available models	No	gpt-4o
system-prompt	Text that will be used as system prompt	No	"You are a helpful assistant."
system-prompt-file	Path to a file containing the system prompt	No	-
response-schema-file	Path to a file containing the response JSON Schema for structured outputs	No	-
vars	YAML-formatted variables for Nunjucks variable substitution in prompts	No	-

* Either prompt or prompt-file must be provided

Outputs 📤

Output	Description
text	The AI's response to your prompt