Frequently asked questions

Is this just a prompt collection?

No. A prompt is a single block of text you paste somewhere. Each recipe here is a structured bundle: a strict contract for inputs and outputs, an operational checklist, complete examples, safety gates, exporters for different agents, and a verification status that is tracked separately from the content.

Does "supported" mean an agent actually ran the workflow?

No. Supported means the exporter produces a format that has been confirmed against official documentation. Whether an agent executed a workflow, and whether the result was any good, are separate stages with their own evidence. The compatibility matrix shows all of them side by side.

Does the CLI use the network?

No normal command does. Cloning the repository and installing dependencies are setup steps you run yourself; after that, everything works offline.

Can a recipe execute code on my machine?

No. Recipes are YAML and Markdown data, and the CLI never executes their content. A recipe's text may suggest commands for you or your agent to run, which is exactly why you should read a workflow before using it.

How do I run the CLI from npm?

Install @kauanpolydoro/agentic-workflows globally and use awf for regular interactive use. Use npx --yes @kauanpolydoro/agentic-workflows@latest list for an occasional run, or install the package as a project dependency when you need to pin its version.

Can I edit a file after installing it?

Yes. The CLI notices the changed hash during update and remove and preserves your edited file unless you explicitly pass --force.