hugging-face-model-trainer

SAFE max severity MEDIUMSkillSpector Do not install 100 registry huggingface last scanned 2026-02-07 1 scan

Verdict safe, but 14 findings (14 medium), mainly unauthorized tool use, command injection, transitive trust abuse. Most recent scan 2026-02-07.

Findings14

Critical0

High0

Medium14

Low0

Analyzersstatic llm meta

Second scanner: NVIDIA SkillSpector

VerdictDo not install 100

Score100 / 100

Issues26

CategoriesPrivilege Escalation Dangerous Code Execution Tool Misuse

Version2.12.0 (static)

The two scanners disagree: Cisco found nothing and SkillSpector says do not install. One of them has a blind spot here; read the skill before trusting either. How the two scanners compare across the whole corpus: Two scanners, one corpus.

What this skill does

This skill should be used when users want to train or fine-tune language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV scripts with PEP 723 format, dataset preparation and validation, hardware selection, cost estimation, Trackio monitoring, Hub authentication, and model persistence. Should be invoked for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging…

Package: 15 files, 128 KB
SKILL.md: 3291 words, 12+ sections
Code: Python (6)
License: Complete terms in LICENSE.txt

SKILL.md covers: TRL Training on Hugging Face Jobs · Overview · etc. · When to Use This Skill · Key Directives · Local Script Dependencies · Prerequisites Checklist · ✅ **Account & Authentication** · ✅ **Dataset Requirements** · ⚠️ **Critical Settings** · Asynchronous Job Guidelines · Action Required

Findings

MEDIUM Attempting to install system packages with elevated privileges Unauthorized Tool Usestatic

Pattern detected: sudo apt-get install

scripts/convert_to_gguf.py:27

- Ubuntu/Debian: sudo apt-get update && sudo apt-get install -y build-essential cmake

Remediation: Use user-level installs without sudo. Document if system install is necessary

MEDIUM Attempting to install system packages with elevated privileges Unauthorized Tool Usestatic

Pattern detected: sudo yum install

scripts/convert_to_gguf.py:28

- RHEL/CentOS: sudo yum groupinstall -y "Development Tools" && sudo yum install -y cmake

Remediation: Use user-level installs without sudo. Document if system install is necessary

MEDIUM Attempting to install system packages with elevated privileges Unauthorized Tool Usestatic

Pattern detected: sudo apt-get install

scripts/convert_to_gguf.py:57

print("     Ubuntu/Debian: sudo apt-get install git")

Remediation: Use user-level installs without sudo. Document if system install is necessary

MEDIUM Attempting to install system packages with elevated privileges Unauthorized Tool Usestatic

Pattern detected: sudo yum install

scripts/convert_to_gguf.py:58

print("     RHEL/CentOS: sudo yum install git")

Remediation: Use user-level installs without sudo. Document if system install is necessary

MEDIUM Attempting to install system packages with elevated privileges Unauthorized Tool Usestatic

Pattern detected: sudo apt-get install

scripts/convert_to_gguf.py:68

print("     Ubuntu/Debian: sudo apt-get install build-essential cmake")

Remediation: Use user-level installs without sudo. Document if system install is necessary

MEDIUM Attempting to install system packages with elevated privileges Unauthorized Tool Usestatic

Pattern detected: sudo yum install

scripts/convert_to_gguf.py:69

print("     RHEL/CentOS: sudo yum groupinstall 'Development Tools' && sudo yum install cmake")

Remediation: Use user-level installs without sudo. Document if system install is necessary

MEDIUM HTTP client library imports that enable external communication Data Exfiltrationstatic

Pattern detected: import urllib.request

scripts/dataset_inspector.py:24

import urllib.request

Remediation: Ensure network access is necessary and documented. Review all URLs

MEDIUM Undeclared network usage Unauthorized Tool Usestatic

Skill code uses network libraries but doesn't declare network requirement

Remediation: Declare network usage in compatibility field or remove network calls

MEDIUM Potential description-behavior mismatch Social Engineeringstatic

Skill performs actions not reflected in its description

SKILL.md

Remediation: Ensure description accurately reflects all skill capabilities

MEDIUM TRANSITIVE TRUST ABUSE detected by YARA Transitive Trust Abusestatic

Detects skills that delegate trust to untrusted external content: run scripts

SKILL.md:47

To run scripts locally (like `estimate_cost.py`), install dependencies:

Remediation: Review and remove transitive trust abuse pattern

MEDIUM TRANSITIVE TRUST ABUSE detected by YARA Transitive Trust Abusestatic

Detects skills that delegate trust to untrusted external content: run scripts

SKILL.md:416

--dataset trl-lib/Capybara \

Remediation: Review and remove transitive trust abuse pattern

MEDIUM CODE EXECUTION detected by YARA Command Injectionstatic

Detects dangerous code execution patterns in agent skills (Python/Bash): subprocess.run(

scripts/convert_to_gguf.py:63

has_make = subprocess.run(["which", "make"], capture_output=True).returncode == 0

Remediation: Review and remove code execution pattern

MEDIUM CODE EXECUTION detected by YARA Command Injectionstatic

Detects dangerous code execution patterns in agent skills (Python/Bash): subprocess.run(

scripts/convert_to_gguf.py:64

has_cmake = subprocess.run(["which", "cmake"], capture_output=True).returncode == 0

Remediation: Review and remove code execution pattern

MEDIUM CODE EXECUTION detected by YARA Command Injectionstatic

Detects dangerous code execution patterns in agent skills (Python/Bash): subprocess.run(

scripts/convert_to_gguf.py:81

result = subprocess.run(

Remediation: Review and remove code execution pattern

What is in the package

File	Kind	Size
`SKILL.md`	doc	27 KB
`references/gguf_conversion.md`	doc	9.6 KB
`references/hardware_guide.md`	doc	6.6 KB
`references/hub_saving.md`	doc	8.3 KB
`references/reliability_principles.md`	doc	11 KB
`references/trackio_guide.md`	doc	5.7 KB
`references/training_methods.md`	doc	4.9 KB
`references/training_patterns.md`	doc	6.0 KB
`references/troubleshooting.md`	doc	8.7 KB
`scripts/convert_to_gguf.py`	script	12 KB
`scripts/dataset_inspector.py`	script	15 KB
`scripts/estimate_cost.py`	script	4.7 KB
`scripts/train_dpo_example.py`	script	3.0 KB
`scripts/train_grpo_example.py`	script	2.3 KB
`scripts/train_sft_example.py`	script	3.3 KB

External hosts referenced

huggingface.co github.com raw.githubusercontent.com hf.co datasets-server.huggingface.co discuss.huggingface.co docs.astral.sh gist.githubusercontent.com lmstudio.ai ollama.ai

Hostnames that appear in the skill's files. Referencing a host is not a finding; compare against the findings above and the skill's stated purpose.

Environment variables referenced

HF_TOKEN ADAPTER_MODEL BASE_MODEL HF_USERNAME

Package contents recorded 2026-10-22 at commit 2afeee6dc727. The scan above ran against the version current at scan time; the two can differ if the skill has since been updated.

What we checked

Static analyzer. Pattern rules over SKILL.md and bundled files: known-bad commands, exfiltration URLs, secrets, obfuscation.
Llm analyzer. A language model reads the skill and reports intent-level issues the rules cannot express.
Meta analyzer. A second model pass that validates, deduplicates and re-scores the other analyzers' findings.

What we did not check

Runtime behaviour. The skill was not executed; findings come from reading its files.
Anything fetched at install or run time, including remote scripts, package dependencies, and URLs the skill tells the agent to open.
Content the analyzers cannot read: images, binaries, and some non-English text.
Known scanner bypass classes. See the methodology and limitations page.

Sources

Repository: https://github.com/huggingface/skills
Raw scan result: skills-hugging-face-model-trainer-scan.json
Badge for your README: ![skillscan.io](https://skillscan.io/badge/huggingface-hugging-face-model-trainer.svg)
Request a re-scan