run-train Skill for Claude: Conservative Training Runs With Evidence

run-train makes your agent run a chosen training command conservatively and record the command, config, seed, logs, checkpoints and metrics.

Installs: 344.4K; on skills.sh
Stars: 486; on GitHub
Files: 4; besides SKILL.md
Updated: Sep 2026; last repo commit

run-train executes a training command you or an earlier step already chose. Modes are startup verification, short-run verification, full kickoff and resume. It records command, config, seed, log, checkpoint, status and metric evidence in train_outputs/.

It does not choose the research goal, set up the environment or branch into sweeps. Give it the goal, a runnable command, environment assumptions and a run mode. It reads shared files from ai-research-reproduction, so install the whole set.

Good for

Researchers who want a training startup or short run verified before a long one
ML engineers resuming training with a clean evidence trail
Teams that need partial, blocked and resumed states recorded clearly

Install run-train

Claude Code

Copy this prompt.

Install the run-train skill in Claude Code. Run this command:

npx skills add lllllllama/RigorPilot-Skills --all

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

This skill reads shared guideline files from the ai-research-reproduction skill, so install the whole set rather than this folder alone. The README says the main reproduction skill works alone and that "all skills" is the choice for the companion and leaf skills.

From the project README. Source.

What's in the skill

When it loads

Rigor Train skill for deep learning research repositories. Use when a documented or selected training command should be run conservatively for startup verification, short-run verification, full kickoff, or resume, with command, config, seed, log, checkpoint, status, and metric evidence written to standardized `train_outputs/`. Do not use for environment setup, exploratory sweeps, speculative idea implementation, or end-to-end orchestration.

Sections in SKILL.md

When to apply
When not to apply
Clear boundaries
Input expectations
Output expectations
Notes

Key facts

Author: lllllllama (owner of the lllllllama/rigorpilot-skills repo)
License: MIT
Installs on skills.sh: 344,449 as of 2026-11-02
GitHub stars: 486 for the whole lllllllama/rigorpilot-skills repo
skills.sh audits: Gen Agent Trust Hub warn, Socket pass, Snyk pass
What is inside: 4 files besides SKILL.md (2 in scripts/, 1 in agents/, 1 in references/)
Needs: The other RigorPilot skills for shared reference files; Python for the bundled run_training.py helper; your own training environment
Last commit to the repo: 2026-09-23

What it includes

Startup check, short run, full kickoff, resume
You pick the mode and the command. The skill runs it conservatively.
Try: “Use run-train for a startup verification of `python train.py --config base.yaml`.”
Runs code · Run modes
Bundled Python script named in SKILL.md
SKILL.md names scripts/run_training.py and scripts/write_outputs.py in its notes but does not say what they do.
Runs code · scripts/run_training.py
Partial, blocked, resumed and kicked-off
SKILL.md says to record partial, blocked, resumed and kicked-off states clearly.
Write · Record states
Configs, seeds, checkpoints, logs, metrics
Saved when available, with runtime assumptions.
Write · Keep reproducibility context
SUMMARY, COMMANDS, LOG, changelog, comparability, status
Six files: SUMMARY.md, COMMANDS.md, LOG.md, SCIENTIFIC_CHANGELOG.md, COMPARABILITY_REPORT.md and status.json.
Write · Write train_outputs/

From the official docs

Permissions

Medium risk

The agent starts real training runs on your machine. Try a startup or short run before a full kickoff.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It starts real training on your machine
SKILL.md covers startup verification, short-run verification, full kickoff and resume. Pick startup or short-run mode first.
It runs the command you give it
SKILL.md expects a runnable training command from you. The README says approved commands can access the host and network, so read it before you approve it.

Technical details

Scopes	Runs the training command you choose Bundles two Python scripts (run_training.py, write_outputs.py) that SKILL.md names Writes train_outputs/ plus the checkpoints and logs your training writes

Questions

What does run-train do?

It runs a selected training command carefully and writes standardized evidence: command, config, seed, logs, checkpoints, status and metrics.

How do I install it?

Use the install box at the top of this page, which installs the whole RigorPilot set. The skill reads shared files from ai-research-reproduction, so installing it alone is not enough. The README says the installer needs Node.js/npm and that skills@1.5.26 requires Node 22.20.0 or newer.

Does it need an API key?

No key is named in SKILL.md.

What does it need from me?

A training goal, a runnable training command, environment and asset assumptions, and a run mode such as startup verification or resume.

When does it not apply?

Not for environment setup, exploratory sweeps, speculative idea implementation or end-to-end orchestration, per its SKILL.md description.

Is it safe?

It executes a training command you or an earlier step selected, so read that command first. The README says RigorPilot is local execution, not an OS sandbox, so use trusted repos. The folder also bundles two Python scripts.

Information

updated Nov 2, 2026

Repository: lllllllama/rigorpilot-skills
Folder: skills/run-train

lllllllama/rigorpilot-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 3

RigorPilot Skills README · github.com
run-train SKILL.md · github.com
skills.sh listing and audits · skills.sh