Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.osmosis.ai/llms.txt

Use this file to discover all available pages before exploring further.

Use this path when you already have a task, dataset, or existing agent code and want an AI coding agent to help turn it into a runnable Osmosis rollout. If you are new to Osmosis and want the shortest copy-paste path, start with Run the Multiply Example instead.
This guide assumes you have completed Onboarding: your workspace repository is cloned, the CLI is installed and authenticated, and your AI coding environment is open in the workspace directory.

​
What Workspace Skills Do

Platform-created workspace repositories include project-local Agent Skills under .agents/skills/. Agents that support the open Agent Skills format can use those skills to move through the same local loop an experienced Osmosis user would follow: 
plan from dataset -> create rollout -> run local eval -> debug failures -> prepare training
It is not a replacement for the CLI. The agent still uses the Osmosis CLI as the source of truth for workspace checks, dataset validation, evals, and training preflight.

​
When to Use This Path

Use this path whenRun the multiply example when
You already know the task you want to train onYou want proof that the platform works end to end
You have sample data or a platform datasetYou do not want to design a dataset yet
You want the agent to create or adapt rollout codeYou want to copy commands without making product choices
You are comfortable inspecting generated code and eval outputYou are still learning the Osmosis workflow

​
Use the Skills in Your Workspace Repository

Open your platform-created workspace repository in your AI coding environment. The repository includes the workspace contract and Agent Skills alongside rollout code, configs, and data: 
repository/
├── .agents/
│   └── skills/
├── .claude/
│   └── skills/
├── rollouts/
├── configs/
│   ├── eval/
│   └── training/
├── data/
├── AGENTS.md
├── CLAUDE.md
└── pyproject.toml
AGENTS.md contains the always-loaded workspace contract. .agents/skills/ contains the canonical workflow skills, and .claude/skills/<skill-name> exposes the same skills to Claude Code through symlinks back to .agents/skills/.

​
Start in a Workspace Repository

The skills assume this repository layout for source files: 
repository/
├── rollouts/
├── configs/
│   ├── eval/
│   └── training/
├── data/
└── pyproject.toml
Before asking the agent to write rollout code, confirm that the CLI can resolve the workspace: 
osmosis doctor
osmosis auth whoami

​
Ask the Agent to Plan from the Dataset

Start by describing your task domain and asking the agent to design the first dataset and eval loop. A useful first prompt is: 
I want to train a model for <my task domain>. Propose a good method to generate a small synthetic dataset under data/, create or adapt an initial rollout and grader for the task, run a quick local eval baseline, inspect the results, and recommend how we should iterate before submitting training.
The workspace skills should guide the agent to:
1

Plan training

Inspect data/, existing rollouts, and workspace config. The agent should settle the dataset schema before writing rollout code.
2

Create or adapt a rollout

Write the smallest AgentWorkflow and Grader that can load, run, and score samples. Generated files should stay under rollouts/, configs/eval/, configs/training/, and data/.
3

Run a local eval

Use local eval as the quality gate:
osmosis eval run configs/eval/<name>.toml --limit 1 --fresh
4

Debug until the eval is clean

Fix loading, dataset, grader, dependency, and reward issues before training. A passing local eval is the handoff point from creation to training readiness.
5

Prepare training

Once the rollout is validated, let the agent inspect the training config and run submit-time preflight. Submit only when you are ready to start a platform training run.
Do not skip local eval. osmosis train submit should be the step after the rollout loads, runs, and grades samples locally.

​
Workspace Skills

The workspace skills are organized around rollout creation stages:
SkillPurpose
plan-trainingTurn a task idea or dataset into a concrete local experiment plan
create-rolloutsCreate or adapt rollout code, graders, entrypoints, and baseline eval configs
evaluate-rolloutsRun local evals, compare baselines, and inspect failures
debug-rolloutsDiagnose eval, config, dataset, dependency, or preflight failures
submit-trainingPrepare a training config, submit training, and check training status
You usually do not need to invoke these skills by name. Describe the outcome you want, and the agent should apply the right stage.

​
Next Steps

Rollout Overview

Understand the AgentWorkflow and Grader contract behind generated rollout code.

Local Evaluation

Validate rollouts before submitting training.

Git Sync

Push rollout changes and let the platform sync the code version used for training.

Training Runs

Submit and monitor training after local eval passes.