distil labs Agent Skill

---
name: distil-cli
metadata:
  version: "14.1.0"
description: >
  Use when building or training a model on the distil labs platform end to end: preparing a
  dataset (config.yaml, job_description.json, traces, train and test rows), running or
  analyzing any pipeline stage (building a test set or a train set from traces by relabelling
  and synthetic data generation, teacher evaluation, model training, deployment), iterating
  when a teacher score or student score is too low, or deciding what to run next after a
  stage completes, or iterating on a model on its own until it reaches a target
  (autoresearch). Activate for phrasings like "build a model for X", "train a student model",
  "build a test set from my traces", "run a teacher eval", "generate synthetic data", "the
  score is low, what now", "add more training data", or "keep iterating until the model is as
  good as you can get it". Stages run through the distil CLI or
  the distil labs API. This skill owns the model-building logic on top of it. Also activate for
  distil labs lookups that are not a full build: distil CLI commands (distil auth,
  distil traces, distil dataset, distil teacher-evaluation, distil slm, distil deployment,
  distil inference-endpoint, distil api-keys, distil access-token), the distil labs REST API,
  config.yaml parameters, supported student and teacher models, what an evaluation metric
  means, or moving a seed dataset or training dataset into a Dataset.
---

# Building Models

Building task-specific small language models on the distil labs platform: from production
traffic or files on disk, through a test set, teacher evaluation and a train set, to a
finetuned, evaluated student served behind an inference endpoint.

Before the first stage, install the `distil` CLI and sign in per
`references/execution/README.md` § Set up the CLI, then settle the backend per § Choose the
backend.

## How the Skill Is Organised

The skill separates what to do from how to run it. Each fact has exactly one owning page, and
every other page links to it.

- **Stages** (`stages/`): one unit of pipeline work each, directly invocable. A stage file is a
  step-by-step procedure: what it needs, what to confirm with the user, how to run it, how to