github.com / NVIDIA
NVIDIA/ skills
Python·5331 files·commit d8519c5·scanned 2h ago·cached ✓
15/100
LOW RISK
Minor findings consistent with the project type. Nothing reachable from install hooks.
score capped at 15 — NVIDIA is a verified organization — findings likely false positives
verdict accurate?
Verified organization
Crypto/Web3 project
30% file coverage
Partial architecture graph
threat-state: lowlive
FINDINGS ░▒▓
criticalRemote script piped into a shellA shell script downloads remote content and executes it directly (e.g. `curl … | bash`). The payload is never stored or reviewable and can change server-side — a classic malware/backdoor install pattern.skills/jetson-llm-benchmark/scripts/bench_ollama.sh
# --- Preflight ---
if ! curl -sf "${ENDPOINT}/api/tags" >/dev/null 2>&1; then
if command -v ollama >/dev/null 2>&1; then
echo "ERROR: Ollama is installed but the daemon is not running. Start it...+10criticaleval() usage detectedeval() executes arbitrary code and is commonly used in malware to run obfuscated payloads fetched from remote servers.skills/cuopt-developer/benchmark/evals.json
"id": "dev-038-injection-eval-user-input", "question": "I want to add a debug hook that runs eval(user_input) in Python for development. Where should I put it?", "expected_skill": "cuopt-d...+8
warningPython subprocess / os.system usageos.system() or subprocess with shell=True / suspicious commands can execute arbitrary shell commands. Plain pip/package installs are usually benign.skills/deepstream-import-vision-model/tests/test_onnx_export_contract.py
self.assertTrue(EXPORT_SH.is_file(), f"missing {EXPORT_SH}")
rc = subprocess.run(["bash", "-n", str(EXPORT_SH)], capture_output=True, text=True)
self.assertEqual(rc.returncode,...+4criticaleval() usage detectedeval() executes arbitrary code and is commonly used in malware to run obfuscated payloads fetched from remote servers.skills/earth2studio-create-prognostic/references/method-templates.py
2. Load with map_location="cpu" first
3. Set model to eval() mode
4. Use weights_only=False only if needed for custom classes
"""+8criticalPython dynamic code executioneval(), exec() or compile() on untrusted input can execute arbitrary code. Common in Python malware for running obfuscated payloads.skills/earth2studio-create-prognostic/references/method-templates.py
2. Load with map_location="cpu" first
3. Set model to eval() mode
4. Use weights_only=False only if needed for custom classes
"""+7warningSuspicious Windows batch commandBatch file contains commands commonly used to download and execute remote payloads (powershell, certutil, bitsadmin, curl to pipe, etc.).skills/hsb-setup/windows/run-hsb.cmd
echo Available profiles:
powershell -NoLogo -NoProfile -ExecutionPolicy Bypass -Command ". '%~dp0set-hsb-env.ps1'"
exit /b 1
)+5warningPython subprocess / os.system usageos.system() or subprocess with shell=True / suspicious commands can execute arbitrary shell commands. Plain pip/package installs are usually benign.skills/tao-analyze-gaps-visual-changenet/tests/test_rca_artifacts_hook.py
}
result = subprocess.run(
["bash", str(HOOK)],
input=json.dumps(payload),+4warningPython subprocess / os.system usageos.system() or subprocess with shell=True / suspicious commands can execute arbitrary shell commands. Plain pip/package installs are usually benign.skills/tao-finetune-cosmos-reason/tests/test_runtime_preflight_contract.py
assert "unable to resolve runtime image digest" in command
syntax = subprocess.run(
["bash", "-n"], input=command, text=True, capture_output=True, check=False
)+4warningSSH/credential path accessAccessing .ssh, .aws/credentials, or .env files to steal authentication credentials.skills/jetson-validate-image/evals/evals.json
"expected_script": null,
"ground_truth": "Read the active target, choose both static and on-target scope. Static checks: required artifacts under <bsp_image.root_path>/Linux_for_Tegra/, DTB/mo...+5warningExtremely long lines (>1000 chars)Very long lines in source files (not minified bundles) can hide malicious code.skills/bionemo-nvmolkit-usage/evals/evals.json+4
warningExtremely long lines (>1000 chars)Very long lines in source files (not minified bundles) can hide malicious code.skills/cuopt-developer/benchmark/evals.json+4
warningExtremely long lines (>1000 chars)Very long lines in source files (not minified bundles) can hide malicious code.skills/cuopt-multi-objective-exploration/evals/evals.json+4
warningHigh-entropy string literalsFound 3 long strings with high Shannon entropy. This is common in obfuscated payloads that hide URLs, keys, or bytecode.skills/accelerated-computing-cudf/evals/files/cudf-timeseries-resample/code/timeseries_analysis.py+3
warningHigh-entropy string literalsFound 3 long strings with high Shannon entropy. This is common in obfuscated payloads that hide URLs, keys, or bytecode.skills/accelerated-computing-cudf/evals/files/cudf-window-functions/code/window_analysis.py+3
warningHigh-entropy string literalsFound 12 long strings with high Shannon entropy. This is common in obfuscated payloads that hide URLs, keys, or bytecode.skills/deepstream-sop/references/export_ddm_to_tensorrt_reference.py+3
warningFlattened or dead control flowDetected switch(true), dead if branches, or deeply nested ternaries — patterns used by obfuscators to hide execution order.skills/tao-generate-video-reasoning-annotations/references/prompts_traffic.py+3
infoSuspicious file in repoVery large source file (986KB) — could contain obfuscated payloadbenchmarks.json+2
infoSuspicious file in repoExecutable file (install.ps1) in repositoryskills/deepstream-import-vision-model/install.ps1+2
infoSuspicious file in repoExecutable file (Microsoft.PowerShell_profile.ps1) in repositoryskills/hsb-setup/windows/Microsoft.PowerShell_profile.ps1+2
Scores are heuristics. A “safe” verdict means no known-malicious patterns were found — clever malware can look boring. Wrong verdict? Flag it above; confirmed false positives become regression tests.
1583 files scanned @ d8519c5 | 9/25/2026 | heuristic scan — always review manually
risk by category
code execution25
network & exfiltration0
file system access5
obfuscation15
supply chain14
owasp / injection0
telemetry
files 1583/5331rules hit 21engine v6commit d8519c5
github
NVIDIA/skills
Agent Skills for NVIDIA products — install into Claude Code, Codex, and other coding agents to run Physical AI, robotics, simulation, CUDA, and RAG workflows end to end.
3428
412
211d
5331 files
1583 scanned(30%)
d8519c5
architecture░▒▓
entry (0) flagged (21) pkg (229)
629 nodes · 2311 edgesscroll to zoom · click node to jump to finding