Skills run-train

📦

run-train

Name: run-train
Author: lllllllama

Content revision r2 High Risk ⚙️ External commands📁 Filesystem access

Run Reproducible Deep Learning Training

Training runs often lack consistent evidence and clear status. This skill executes a selected command and records logs, metrics, checkpoints, changes, and outcomes.

Supports: Claude Codex Code(CC)

⚠️ 38 Poor

Stop for confirmation before installing.

Review the plan and obtain explicit user consent before changing files.

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "run-train" from https://skillstore.io/skills/lllllllama-run-train.md and its manifest at https://skillstore.io/api/skills/lllllllama-run-train/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/lllllllama-run-train.md Signed manifest GET /api/skills/lllllllama-run-train/manifest Signed lockfile GET /api/skills/lllllllama-run-train/lockfile

Test it

Using "run-train". Startup verification with a two-minute monitoring window.

Expected outcome:

Status: partial because the monitoring window elapsed after step 24.
Evidence: validation accuracy reached 0.71 and a step 24 checkpoint appeared.
Repository changes: one new checkpoint and one updated training log.

Using "run-train". Resume a selected run from an existing checkpoint.

Expected outcome:

Status: success because the resumed process completed without an error.
Progress: training continued from epoch 4 to epoch 5.
Evidence: the report records the checkpoint, latest metric, command outcome, and touched files.

Security Audit

High Risk

v7 • 8/8/2026 Open versioned report

Most detections in SKILL.md are Markdown code spans or fixed documentation paths and are false positives. The runner intentionally executes a user-provided training command. The output writer also imports and executes a module outside the audited skill package, requiring stronger trust controls.

Files scanned

509

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (1)

High

scripts/write_outputs.py:11-16

Unaudited External Module Execution

The output wrapper resolves a shared Python file three parent directories above itself and executes that module. This crosses the audited package boundary and permits local code substitution.

The code directly constructs an ancestor-relative module path and calls exec_module. The executed module is not included among the audited skill files.

Capability review items (1)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

High

scripts/run_training.py:177

Python subprocess.run

result = subprocess.run(

The command option is parsed into an argument list and executed directly in the selected repository. Any supplied executable can therefore run with the agent's permissions.

Risk Factors

⚙️ External commands (12)

scripts/run_training.py:81 scripts/run_training.py:177 SKILL.md:3 SKILL.md:8 SKILL.md:12 SKILL.md:47 SKILL.md:48 SKILL.md:49 SKILL.md:50 SKILL.md:51 SKILL.md:52 SKILL.md:56

📁 Filesystem access (2)

SKILL.md:12 SKILL.md:56

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/lllllllama-run-train/audits/7?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/lllllllama-run-train/security.svg)](https://skillstore.io/skills/lllllllama-run-train?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/lllllllama-run-train?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/lllllllama-run-train/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/lllllllama-run-train.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

lllllllama. (2026). run-train security audit report (audit version 7) [Author version unspecified]. Skillstore. https://skillstore.io/skills/lllllllama-run-train/audits/7

BibTeX citation

@techreport{lllllllama-lllllllama-run-train-2026,
  author = {lllllllama},
  title = {run-train security audit report (audit version 7)},
  institution = {Skillstore},
  year = {2026},
  number = {7},
  url = {https://skillstore.io/skills/lllllllama-run-train/audits/7},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "run-train security audit report (audit version 7)"
version: "unspecified"
type: report
authors:
  - name: "lllllllama"
date-released: "2026-08-08"
url: "https://skillstore.io/skills/lllllllama-run-train/audits/7"
identifiers:
  - type: other
    value: "skillstore:lllllllama-run-train:audit:7"
    description: "Skillstore immutable audit report identifier"

Skillstore Score

Why this score Evidence Confidence: Medium

Architecture

Maintainability

Content

Community

Spec Compliance

What You Can Build

Verify Training Startup

Run a selected command briefly and confirm that steps, metrics, and checkpoints begin appearing.

Document Reproduction Runs

Capture command, dataset, checkpoint, status, and repository changes for a controlled reproduction attempt.

Monitor Resume Operations

Resume from a known checkpoint and record whether progress continues within a bounded monitoring window.

Try These Prompts

Check Training Startup

Verify startup in [repository] using [command]. Monitor for [seconds] seconds and report steps, metrics, checkpoints, and blockers.

Run a Short Verification

Run [command] in [repository] for [seconds] seconds. Record dataset assumptions, changed files, latest step, metric, and checkpoint.

Resume From a Checkpoint

Resume [command] from [checkpoint] in [repository]. Confirm progress, capture logs, and distinguish a successful resume from partial monitoring.

Kick Off a Documented Run

Kick off [command] in [repository] using [dataset]. Preserve seed and configuration context, monitor for [seconds] seconds, and produce all evidence files.

Best Practices

Review and confirm the exact command, repository, dataset, and checkpoint before execution.
Use a short monitoring window before committing resources to a full run.
Verify parsed metrics and checkpoint paths against the original logs before drawing conclusions.

Avoid

Do not use this skill to explore unknown commands or autonomous experiment variants.
Do not treat startup success as evidence of complete reproduction.
Do not execute commands from untrusted repositories without isolation and explicit review.