Minimal Run and Audit Skill: Log Repro Runs

Captures evidence from a smoke test or documented inference or evaluation command and writes standard repro_outputs/ files.

Installs: 484.1K; on skills.sh
Stars: 486; on GitHub
Files: 4; besides SKILL.md
Updated: Sep 2026; last repo commit

minimal-run-and-audit is the run step of README-first deep learning reproduction. It takes a selected short command, such as a smoke test or documented inference or evaluation run, and writes normalized evidence: a result summary, repro_outputs/ files, SCIENTIFIC_CHANGELOG.md, COMPARABILITY_REPORT.md and PATCHES.md when repo files changed.

It marks each outcome as verified, partial or blocked, and must not hide changes that alter evaluation, preprocessing, checkpoints or metrics. It is not for training, first-time repo scanning, environment setup or choosing the target. It reads shared principles from the sibling ai-research-reproduction skill and bundles run_command.py and write_outputs.py.

Good for

Researchers who need auditable evidence of a reproduction run
ML engineers verifying a repo's smoke test or eval command
Users of the RigorPilot main skill

Install minimal-run-and-audit

Claude Code

Copy this prompt.

Install the minimal-run-and-audit 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 22.20+.

The README says the main skill, ai-research-reproduction, works alone, and to choose all skills for the companion entrypoints. SKILL.md reads shared files from the ai-research-reproduction skill, so install the set. The README tested the installer with skills@1.5.26 and Node 22.20.0, which it says is the minimum.

From the project README. Source.

What's in the skill

When it loads

Rigor Run skill for README-first deep learning repo reproduction. Use when the task is specifically to capture or normalize evidence from the selected smoke test or documented inference or evaluation command and write standardized `repro_outputs/` files, including patch notes when repository files changed. Do not use for training execution, initial repo intake, generic environment setup, paper lookup, target selection, hidden scientific-meaning changes, or end-to-end orchestration by itself.

Sections in SKILL.md

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

Key facts

Author: lllllllama (RigorPilot-Skills)
Installs on skills.sh: 484,123 (Nov 2, 2026)
GitHub stars: 486 on lllllllama/rigorpilot-skills (Nov 2, 2026)
License: MIT
Last repo commit: Sep 23, 2026
What installs: SKILL.md plus 4 bundled files: agents/openai.yaml, references/reporting-policy.md, scripts/run_command.py, scripts/write_outputs.py
skills.sh audits: Gen Agent Trust Hub: pass, Socket: warn, Snyk: pass
Part of: RigorPilot-Skills; reads shared principles from the ai-research-reproduction skill
Runs: scripts/run_command.py and scripts/write_outputs.py
Writes: repro_outputs/, SCIENTIFIC_CHANGELOG.md, COMPARABILITY_REPORT.md and PATCHES.md when repo files changed

What it includes

Executes a smoke test or documented inference or eval command
Short, non-training checks only. Uses scripts/run_command.py.
Runs code · Run the selected command
Writes standardized evidence files
Uses scripts/write_outputs.py.
Files · Write repro_outputs/
Records changes in scientific meaning and evidence status
Must not hide changes to evaluation, preprocessing, checkpoints or metrics.
Files · SCIENTIFIC_CHANGELOG.md
Compares README, paper and baseline comparability
Part of the standard outputs.
Files · COMPARABILITY_REPORT.md
Lists repo files that changed
Written only when repository files changed.
Files · PATCHES.md
Marks results verified, partial or blocked
Keeps the three states distinct.
Label outcomes

From the official docs

Permissions

Medium risk

It runs a command from the target repo and writes evidence files, and the README says this is local execution, not a sandbox. Use trusted repos and review PATCHES.md.

What it can do

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

Watch out for

It runs commands from someone else's repo
The README says RigorPilot is local execution, not an OS sandbox, and approved commands can reach the host and network. Use trusted repos.
It can patch repo files
It writes PATCHES.md when files change. Read that file and the diff before trusting a result.
Its skills.sh listing shows a Socket warn
Open the audit page for details before you install.

Technical details

Talks to	Whatever the repo command itself uses
Scopes	Runs commands from the target repo Writes repro_outputs/ and report files

Questions

What does minimal-run-and-audit do?

It runs or captures a selected smoke test or documented inference or evaluation command and writes standard evidence files under repro_outputs/, including patch notes if repo files changed.

How do I install it?

Follow the install box at the top of the page. It reads files from the sibling ai-research-reproduction skill, so install the whole RigorPilot set.

Does it run training?

No. SKILL.md says it does not own training startup, resume or long-running training state.

What are the output files?

repro_outputs/ files, SCIENTIFIC_CHANGELOG.md, COMPARABILITY_REPORT.md, and PATCHES.md when repo files changed.

Does it pick what to run?

No. It does not choose the overall target; the main skill or you select the command.

Is it safe?

It runs commands from the repo you point at, and the README says approved commands can access the host and network. Use trusted repos, keep runs short, and read PATCHES.md. Its listing shows a Socket warn on skills.sh.

Information

updated Nov 2, 2026

Repository: lllllllama/rigorpilot-skills
Folder: skills/minimal-run-and-audit

lllllllama/rigorpilot-skills skills.sh

Sources

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

Show all 3

RigorPilot-Skills README · github.com
minimal-run-and-audit SKILL.md · github.com
skills.sh listing · skills.sh