A runtime for agentic workflows

Agentic workflows
that keep running.

Define agents and workflows in YAML, run them from your terminal, and resume every execution from persistent state.

flow.yamlAgent step

steps:
  - wait_input: ask_user
    prompt: "What CI, PR, or release task should I prepare?"
    next: ci

  - agent: ci
    system:
      file: "./system.md"
    task: '{{output_value("ask_user").payload.text}}'
    instructions:
      - "skiller-ci"
      - "response-style"
    tools:
      - shell
      - files
    next: ask_user

Agent harness

Give agents a runtime, not just a prompt.

Run agents as explicit workflow steps with configured models, tools, instructions, limits, persistent context, and observable events.

Create your first agent

STUI

Work with agents from your terminal.

Chat with agents, launch workflows, provide human input, and inspect live execution without leaving the terminal.

skiller-stuiAgent session

▪

[tool.hatch.build.targets.sdist]
include = [
  "/README.md",
  "/LICENSE",
  "/pyproject.toml",
  "/packages/skiller/src",
  "/packages/skiller/docs",
  "/apps/tui/src",
  "/apps/agents",
  "/ap...

‹

The Python package configuration now keeps web concerns outside the published distribution. A focused test can make that boundary explicit:

import tomllib
from pathlib import Path


def test_sdist_excludes_website() -> None:
    pyproject = tomllib.loads(
        Path("pyproject.toml").read_text(encoding="utf-8")
    )

    included_paths = pyproject["tool"]["hatch"]["build"]["targets"]["sdist"]["include"]

    assert "/website" not in included_paths
    assert "/.github" not in included_paths
    assert "/.env.development" not in included_paths