hugging-face-model-trainer
Train language models using TRL (Transformer Reinforcement Learning) on fully managed Hugging Face infrastructure. No local GPU setup required—models train on cloud GPUs and results are automatically saved to the Hugging Face Hub.
54
62%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
View guide
High
Do not use without reviewing
Fix and improve this skill with Tessl
tessl review fix ./plugins/AI-Agents-Safe-Coding-Skills-claude/skills/hugging-face-model-trainer/SKILL.mdThe canonical home for this skill is hugging-face-model-trainer in administrakt0r/AI-Agents-Safe-Coding-Skills
Security
1 high severity finding. You should review these findings carefully before considering using this skill.
W007: Insecure credential handling detected in skill instructions.
What this means
The skill handles credentials insecurely by requiring the agent to include secret values verbatim in its generated output. This exposes credentials in the agent’s context and conversation history, creating a risk of data exfiltration.
Why it was flagged
The skill repeatedly instructs the agent to pass an HF token in job submission (e.g., secrets={"HF_TOKEN": "$HF_TOKEN"}) and notes the "$HF_TOKEN" should reference the user's actual token, which forces or encourages embedding a secret value into generated job/config commands and outputs (high exfiltration risk).
Low
Low-risk findings.
2 low severity findings. Worth noting, but not necessarily harmful.
W011: Third-party content exposure detected (indirect prompt injection risk).
What this means
The skill exposes the agent to untrusted, user-generated content from public third-party sources, creating a risk of indirect prompt injection. This includes browsing arbitrary URLs, reading social media posts or forum comments, and analyzing content from unknown websites.
Why it was flagged
At runtime the skill instructs users to submit training scripts via `hf_jobs()` where the `script` parameter may include attacker-controlled code/URLs and the workflow also fetches/uses `dataset_inspector.py` plus datasets provided by the user (e.g., `load_dataset(...)`), meaning outsider-authored free text can be ingested for validation/training.
W012: Unverifiable external dependency detected (runtime URL that controls agent).
What this means
The skill fetches instructions or code from an external URL at runtime, and the fetched content directly controls the agent’s prompts or executes code. This dynamic dependency allows the external source to modify the agent’s behavior without any changes to the skill itself.
Why it was flagged
The skill includes explicit runtime script URLs (e.g., https://huggingface.co/datasets/mcp-tools/skills/raw/main/dataset_inspector.py and other script URLs passed as the "script" parameter) which are fetched by hf_jobs/uv and would execute remote Python code inside the job environment, so they directly control runtime execution.
- Commit
fbdfd63
- Audited
- Security analysis
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.