Skip to content

Latest commit

 

History

History

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 

README.md

Agent Action

agent runs the ChatBotKit CLI agent inside a GitHub Actions job.

This action executes cbk agent via an isolated npm exec invocation of @chatbotkit/cli in the target working directory.

Inputs

  • prompt - Required task prompt passed to cbk agent --prompt
  • tools - Optional tool list passed to cbk agent --tools
  • bot - Optional bot id passed to cbk agent --bot
  • model - Optional model name passed to cbk agent --model
  • dataset - Optional dataset id passed to cbk agent --dataset
  • skillset - Optional skillset id passed to cbk agent --skillset
  • cli-version - Optional CLI version to execute, defaults to latest
  • node-version - Optional Node.js version, defaults to 20
  • working-directory - Optional working directory, defaults to .

The optional CLI-facing inputs are applied in this order: tools, bot, model, dataset, skillset.

Required Environment

Set the secrets and environment variables required by your ChatBotKit CLI configuration. In most cases that means exporting CHATBOTKIT_API_KEY from repository or organization secrets.

Example

jobs:
  agent:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - uses: chatbotkit/github-actions/agent@v0
        with:
          tools: read,write,edit,find,exec
          bot: bot_123
          model: gpt-5
          dataset: dataset_123
          skillset: skillset_123
          prompt: >-
            Read the repository and propose a small documentation improvement.
          cli-version: 1.34.0
          working-directory: .
        env:
          CHATBOTKIT_API_KEY: ${{ secrets.CHATBOTKIT_API_KEY }}

Notes

  • Prefer pinned cli-version values for reproducible workflows.
  • The CLI is executed without a global install, which avoids leaving shared state on self-hosted runners.
  • The action runs with the permissions, checked-out files, and runner access provided by the calling workflow.
  • For centralized orchestration across many repositories, pair this action with a reusable workflow in the same repository.