explore-run Skill for Claude: Bounded Exploratory ML Runs

explore-run makes your agent plan and summarize small exploratory ML runs, ranked by cost, success rate and expected gain.

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

explore-run is the run leaf of the RigorPilot explore lane. With your explicit authorization it plans exploratory runs such as small-subset validation, short-cycle guess-and-check, batch sweeps, idle-GPU search or quick transfer-learning trials. It ranks candidates by cost, success_rate and expected_gain, then by real results once runs finish.

Summaries go to explore_outputs/ and label results as bounded evidence with fair-comparison caveats. You can set variant_axes, subset_sizes, short_run_steps, selection_weights, max_variants and max_short_cycle_runs. It hands real execution to minimal-run-and-audit or run-train, and it reads shared files from ai-research-reproduction, so install the whole set.

Good for

Researchers who want quick small-subset or short-cycle checks before a big run
ML engineers running batch sweeps or using idle GPU time
Anyone who needs run results labeled as bounded evidence

Install explore-run

Claude Code

Copy this prompt.

Install the explore-run 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 Improve / Rigor Explore run leaf skill for bounded exploratory evidence in deep learning research repositories. Use when the researcher explicitly authorizes exploratory runs such as small-subset validation, short-cycle guess-and-check, batch sweeps, idle-GPU search, or quick transfer-learning trials, with fair-comparison caveats and no-overclaim summaries in `explore_outputs/`. Do not use for end-to-end exploration orchestration on top of `current_research`, trusted baseline execution, conservative training verification, default routing, verified SOTA claims, or implicit experimentation.

Sections in SKILL.md

When to apply
When not to apply
Clear boundaries
Ranking Semantics
Variant Spec Hints
Output expectations
Notes

Key facts

Author: lllllllama (owner of the lllllllama/rigorpilot-skills repo)
License: MIT
Installs on skills.sh: 344,425 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 helpers
Last commit to the repo: 2026-09-23

What it includes

variant_axes, subset_sizes, short_run_steps
Defines the grid of candidates and the scale of each trial. primary_metric and metric_goal let later ranking order executed runs.
Try: “Use explore-run: try 3 learning rates on a 5% subset for 200 steps and rank them.”
Read · Variant spec
Cost, success rate and expected gain
Conservative default weights unless you give selection_weights. Budget caps (max_variants, max_short_cycle_runs) prune after scoring.
Read · Rank before running
Plans the variant set
Bundled Python script named in SKILL.md notes. SKILL.md does not say what it does.
Runs code · scripts/plan_variants.py
minimal-run-and-audit or run-train run the commands
explore-run plans and summarizes. It prefers small-subset and short-cycle checks before heavier runs.
Runs code · Hand off execution
Changeset, changelog, comparability, top runs
Writes CHANGESET.md, SCIENTIFIC_CHANGELOG.md, COMPARABILITY_REPORT.md, TOP_RUNS.md and status.json, with no overclaiming.
Write · Write explore_outputs/

From the official docs

Permissions

Medium risk

Exploratory sweeps run on your machine and can use a lot of compute. Set run caps and a budget first.

What it can do

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

Watch out for

Sweeps can run many trials
SKILL.md covers batch sweeps and idle-GPU search, with max_variants and max_short_cycle_runs as caps. Set both, and keep subsets small, before you authorize a sweep.
Runs are exploratory evidence only
SKILL.md labels results as bounded evidence, not verified SOTA.
Runs are not sandboxed
The README says RigorPilot is local execution, not an OS sandbox, and to use trusted repositories.

Technical details

Scopes	Plans and, via run-train or minimal-run-and-audit, starts training or evaluation runs Bundles two Python scripts (plan_variants.py, write_outputs.py) that SKILL.md names Writes explore_outputs/ files

Questions

What does explore-run do?

It plans bounded exploratory runs on an ML repo and summarizes them as ranked, caveated evidence. It does not certify trusted success.

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.

When does it not apply?

Not without explicit exploratory authorization, and not for trusted training, setup, intake or debugging. Use run-train for conservative training and safe-debug for failures.

explore-run or run-train?

explore-run is for experiments that rank candidates. run-train executes one selected training command conservatively and records evidence.

Is it safe?

Real execution is handed to minimal-run-and-audit or run-train, so check what they will run. The README says RigorPilot is local execution, not an OS sandbox. SKILL.md caps a sweep through max_variants and max_short_cycle_runs, so set both.

Information

updated Nov 2, 2026

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

lllllllama/rigorpilot-skills skills.sh

Sources

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

Show all 3

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