Pre-release

This release is a pre-release and may not be stable for production use.

Skiller.run

A runtime for agentic workflows.

Skiller runs agentic flows as durable executions with persistent state, safe resume, and full observability logs.

What It Does

Skiller turns a YAML flow into a durable execution. Define the work, start a run, resume it when new input arrives, and inspect every step along the way.

Define the flow. Combine agents, tools, deterministic steps, and external input in YAML. With an LLM provider and tool policies configured, this two-step flow runs an ongoing agent session:

name: mono
start: ask_user

steps:
  - wait_input: ask_user
    prompt: "What should the agent do?"
    next: mono_agent

  - agent: mono_agent
    system: |
      Complete the user's task and report the result clearly.
    task: '{{output_value("ask_user").payload.text}}'
    tools:
      - shell
      - files
    next: ask_user

Start a durable run. Skiller persists its state, outputs, and runtime events instead of keeping the execution only in memory.

skiller run ./mono.yaml

Resume when input arrives. Waiting state survives process restarts. Send input later and continue the same run from the exact step where it paused.

skiller input receive <run_id> --text "Audit the dependencies" --wait

Observe and manage the run. Check its current state, stream new events, or inspect the complete persisted history.

skiller status <run_id>
skiller observe <run_id>
skiller logs <run_id>

Prefer an interactive experience? Open the TUI to chat with agents, launch flows, and return to previous runs:

skiller

Install

For regular CLI usage, install it with pipx:

pipx install skiller

Usage

STUI: chat and launch runs

Use skiller when you want an interactive terminal UI to chat, launch runs, and manage persisted runs.

skiller

CLI: run and manage flows

Run a packaged, local, or configured flow reference:

skiller run @flows
skiller run @group/name

Direct paths are also supported:

skiller run ./my-flow.yaml
skiller run ~/flows/my-flow.yaml

Inspect and manage runs:

skiller status <run_id>
skiller logs <run_id>
skiller delete <run_id>

Flow Steps

Deterministic:

assign
notify
switch
when

Execution:

agent
shell

Waiting:

wait_input
wait_webhook

Persistence

Skiller persists:

run state
step outputs
runtime log events
waiting states
external event receipts
persisted output bodies

Waiting is persisted, not simulated in memory. A run can stop in WAITING and resume later from stored state.

Project Layout

packages/skiller/src/skiller: runtime and CLI code
apps/agents: bundled agents and authentication flows
packages/skiller/docs: runtime and CLI documentation
packages/skiller/tests: runtime, CLI, and integration tests
apps/tui: Textual UI app

Dependencies

Runtime dependencies are grouped by the capability that uses them:

Area	Dependencies	Used for
Core	`pydantic`, `PyYAML`	Configuration validation and YAML flow loading
LLM providers	`openai`, `boto3`	OpenAI, Codex, and Amazon Bedrock adapters
MCP	`fastmcp`	MCP client connections and tool execution
Webhooks	`fastapi`, `uvicorn`	Local webhook server
Terminal UI	`textual`	Interactive agent chat and run management

Development uses pytest for tests, httpx for HTTP test clients, and ruff for linting. Packages are built with hatchling. Direct dependency constraints live in pyproject.toml; the resolved dependency graph lives in uv.lock.

Documentation

Core guides:

Step references:

Development

Run the main checks:

./.venv/bin/python -m ruff check packages/skiller/src apps/tui/src packages/skiller/tests apps/tui/tests
./.venv/bin/python -m pytest packages/skiller/tests apps/tui/tests
./.venv/bin/python -m build --no-isolation

Manual CLI E2E flows live in packages/skiller/tests/e2e/cli_*.sh. Step-specific flows live in packages/skiller/tests/e2e-steps/<test>/run.sh. Use them when you need to exercise the real CLI path without mixing those checks into the default pytest suite.

License

Apache-2.0. See LICENSE.

Disclaimer

This project is provided "as is", without warranties of any kind. The authors and contributors are not responsible for production incidents, data loss, service interruptions, security issues, regulatory non-compliance, third-party integration failures, or any direct or indirect damages resulting from its use.

Metadata

Release files for skiller 0.1.0b41

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for skiller 0.1.0b41
File	Size	Uploaded
skiller-0.1.0b41.tar.gz	326.8 kB	Sep 30, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for skiller 0.1.0b41
File	Interpreter	ABI	Platform	Reset
skiller-0.1.0b41-py3-none-any.whl 550.3 kB Sep 30, 2026	Python 3	none	any	Details

Total release size: 877.1 kB

Release files / skiller-0.1.0b41.tar.gz

Download URL	skiller-0.1.0b41.tar.gz
Size	326.8 kB
Tags	Source
SHA-256 checksum How to use checksums	`725aca6f2d077a65ee0339e6c91c45db5067b1fc66f6896007bda0c31fd521f9`
BLAKE2b-256 checksum How to use checksums	`1bebaa7ca9c804d56d0392238ee274206405c137951bdf45a408e4a8f0c89103`
Upload date	Sep 30, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/fredycampino/skiller
Publishing commit: github.com/fredycampino/skiller/tree/86c2037f2f0975a8ea70d903dc2f2fce56e38c20

Workflow

Publishing configuration: github.com/fredycampino/skiller/blob/86c2037f2f0975a8ea70d903dc2f2fce56e38c20/.github/workflows/publish.yml
Publishing logs: github.com/fredycampino/skiller/actions/runs/36777564796/attempts/1

Artifact

Subject: skiller-0.1.0b41.tar.gz
SHA-256 checksum: 725aca6f2d077a65ee0339e6c91c45db5067b1fc66f6896007bda0c31fd521f9 Verifying attestations

Release files / skiller-0.1.0b41-py3-none-any.whl

Download URL	skiller-0.1.0b41-py3-none-any.whl
Size	550.3 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`5128b7e8810552e78e56c749e5ed1adf3a690c5428f6ad6e91fad0d1f541fe7d`
BLAKE2b-256 checksum How to use checksums	`49b625bb6975e827aba1cd496a442189dde7712924c2f8d5dec9b213d4b90528`
Upload date	Sep 30, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/fredycampino/skiller
Publishing commit: github.com/fredycampino/skiller/tree/86c2037f2f0975a8ea70d903dc2f2fce56e38c20

Workflow

Publishing configuration: github.com/fredycampino/skiller/blob/86c2037f2f0975a8ea70d903dc2f2fce56e38c20/.github/workflows/publish.yml
Publishing logs: github.com/fredycampino/skiller/actions/runs/36777564796/attempts/1

Artifact

Subject: skiller-0.1.0b41-py3-none-any.whl
SHA-256 checksum: 5128b7e8810552e78e56c749e5ed1adf3a690c5428f6ad6e91fad0d1f541fe7d Verifying attestations