Automodel

A skill to make agents search for better models

Install

npx -y skills add Unlayer-AI/automodel --skill automodel

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Guides the user in discovering a better model structure (e.g. from feature transformations in logistic regression to equation terms within PDEs to neural network layer compositions) out of data. Use when the user asks to create or improve an existing model. Uses an iterative meta/inner agent loop to explore structural model modifications in parallel.

The file declares its own license as Apache-2.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

2.5 KB, as published. Nobody here has run it

automodel

A structured guide to discovering models from data with agents. Focuses on the structure of the model (be it equation terms or neural network layers) rather than just parameter values, using an iterative meta/inner agent loop to explore structural modifications in parallel. Applies to a wide range of tasks, from simple regression to complex physical systems, from automotive to pharmacology.

Phases

This skill is organized into four sequential phases, each with a detailed recipe in assets/phases/:

Phase	Recipe	Entry signal
1 — Setup	`assets/phases/1_setup.md`	No `CONTEXT.md` artifact in project root
2 — Baseline Model	`assets/phases/2_baseline_model.md`	`CONTEXT.md` exists; Some prep work done;
3 — Iterate	`assets/phases/3_iterate.md`	End-to-end pipeline verified for baseline model
4 — Finalize	`assets/phases/4_finalize.md`	User satisfied with validation performance; `CONTEXT.md` points to best model

Each phase updates CONTEXT.md. Use the resulting artifacts as re-entry signals when resuming.

How to start

Copy assets/CHECKLIST.md to the project root (if not already present) and update it as you go.
Detect the current phase using the entry signals above. If a checklist with partial progress is already present, ask the user whether to resume.
Read only the frontmatter of all phase files to confirm which one applies — do not load the full recipes yet:
```
# bash / zsh
bash scripts/read_phases.sh
```
```
# PowerShell
pwsh scripts/read_phases.ps1
```
The frontmatter entry, exit, and description fields are sufficient to identify the applicable phase without loading full recipes into context.
Once the applicable phase is confirmed, load and follow its full recipe (assets/phases/<N>_<name>.md).

References

The user may place relevant papers, code snippets, or other resources in the references folder — check them only when prompted to do so within a particular phase, and refer back / review with the user as needed.