Trust Assessment
huggingface-community-evals received a trust score of 0/100, placing it in the Untrusted category. This skill has significant security findings that require attention before use in production.
SkillShield's automated analysis identified 11 findings: 5 critical, 6 high, 0 medium, and 0 low severity. Key findings include Arbitrary command execution, Dangerous call: subprocess.run(), Unsafe Default Local Sandbox for Code-Executing Evaluations.
The analysis covered 4 layers: Manifest Analysis, Static Code Analysis, Dependency Graph, LLM Behavioral Safety. The Manifest Analysis layer scored lowest at 0/100, indicating areas for improvement.
Last analyzed on August 1, 2026 (commit 87f9ee5b). SkillShield performs automated 4-layer security analysis on AI skills and MCP servers.
Layer Breakdown
Behavioral Risk Signals
Security Findings11
| Severity | Finding | Layer | Location | |
|---|---|---|---|---|
| CRITICAL | Arbitrary command execution Python shell execution (os.system, subprocess) Review all shell execution calls. Ensure commands are static (not built from user input), use absolute paths, and are strictly necessary. Prefer library APIs over shell commands. | Manifest | skills/huggingface-community-evals/scripts/inspect_eval_uv.py:94 | |
| CRITICAL | Arbitrary command execution Python shell execution (os.system, subprocess) Review all shell execution calls. Ensure commands are static (not built from user input), use absolute paths, and are strictly necessary. Prefer library APIs over shell commands. | Manifest | skills/huggingface-community-evals/scripts/inspect_vllm_uv.py:105 | |
| CRITICAL | Arbitrary command execution Python shell execution (os.system, subprocess) Review all shell execution calls. Ensure commands are static (not built from user input), use absolute paths, and are strictly necessary. Prefer library APIs over shell commands. | Manifest | skills/huggingface-community-evals/scripts/inspect_vllm_uv.py:170 | |
| CRITICAL | Arbitrary command execution Python shell execution (os.system, subprocess) Review all shell execution calls. Ensure commands are static (not built from user input), use absolute paths, and are strictly necessary. Prefer library APIs over shell commands. | Manifest | skills/huggingface-community-evals/scripts/lighteval_vllm_uv.py:101 | |
| CRITICAL | Arbitrary command execution Python shell execution (os.system, subprocess) Review all shell execution calls. Ensure commands are static (not built from user input), use absolute paths, and are strictly necessary. Prefer library APIs over shell commands. | Manifest | skills/huggingface-community-evals/scripts/lighteval_vllm_uv.py:165 | |
| HIGH | Dangerous call: subprocess.run() Call to 'subprocess.run()' detected in function 'main'. This can execute arbitrary code. Avoid using dangerous functions like exec/eval/os.system. Use safer alternatives. | Static | skills/huggingface-community-evals/scripts/inspect_eval_uv.py:94 | |
| HIGH | Dangerous call: subprocess.run() Call to 'subprocess.run()' detected in function 'run_inspect_vllm'. This can execute arbitrary code. Avoid using dangerous functions like exec/eval/os.system. Use safer alternatives. | Static | skills/huggingface-community-evals/scripts/inspect_vllm_uv.py:105 | |
| HIGH | Dangerous call: subprocess.run() Call to 'subprocess.run()' detected in function 'run_inspect_hf'. This can execute arbitrary code. Avoid using dangerous functions like exec/eval/os.system. Use safer alternatives. | Static | skills/huggingface-community-evals/scripts/inspect_vllm_uv.py:170 | |
| HIGH | Dangerous call: subprocess.run() Call to 'subprocess.run()' detected in function 'run_lighteval_vllm'. This can execute arbitrary code. Avoid using dangerous functions like exec/eval/os.system. Use safer alternatives. | Static | skills/huggingface-community-evals/scripts/lighteval_vllm_uv.py:101 | |
| HIGH | Dangerous call: subprocess.run() Call to 'subprocess.run()' detected in function 'run_lighteval_accelerate'. This can execute arbitrary code. Avoid using dangerous functions like exec/eval/os.system. Use safer alternatives. | Static | skills/huggingface-community-evals/scripts/lighteval_vllm_uv.py:165 | |
| HIGH | Unsafe Default Local Sandbox for Code-Executing Evaluations The script `inspect_eval_uv.py` configures the `inspect-ai` sandbox backend to default to `local`. When evaluating models on tasks that execute model-generated code (such as `humaneval`, which is listed as a supported task in `SKILL.md`), this default allows untrusted, model-generated code to execute directly on the host system without containerization or isolation. If a model is malicious or compromised via prompt injection, it can execute arbitrary code on the runner's machine. Change the default sandbox to `docker` (or another secure environment) for tasks that execute code, or display a prominent warning to the user when `local` is used with code-executing tasks like `humaneval`. | LLM | scripts/inspect_eval_uv.py:54 |
Scan History
Embed Code
[](https://skillshield.io/report/e11b280d7d0765d1)
Powered by SkillShield