Public scan — anyone with this URL can view this analysis. Sign up to track your own repos privately, run scheduled re-scans, and get AI fix prompts via your dashboard.

15 of your 72 findings came from Repobility's proprietary detections. ✓ Repobility tags below mark them.

Scan timing: clone 1.95s · analysis 4.58s · 1.4 MB · GitHub preflight 673ms

NVIDIA/SkillSpector

https://github.com/NVIDIA/SkillSpector.git · scanned 2026-06-18 08:45 UTC (1 day ago) · 10 languages

95 raw signals (60 security + 35 graph) 12th percentile · Python · small (2-20K LoC)

File as issue Image

UNIFIED Repobility · multi-layer engine · AI coders

Complete repo analysis

Last scanned 1 day ago · v1 · 92 actionable findings from 2 signal sources. 3 repeated signals grouped for readability. Security checks, system graph analysis, and verified AI-agent feedback are merged into one review queue.

JSON

66.7% cov · 33 findings

Score breakdown â 2026-05-18-v5

Component	Sub-score	Weight	Contribution
`structure_score`	85.0	0.15	12.75
`security_score`	50.1	0.25	12.53
`testing_score`	100.0	0.20	20.00
`documentation_score`	100.0	0.15	15.00
`practices_score`	52.0	0.15	7.80
`code_quality`	60.5	0.10	6.05
Overall		1.00	74.1

Severity distribution — click a segment to filter

Active filters: excluding tests × Reset all

Severity: Critical 1 High 37 Medium 20 Low 18 9-Layer: Software Security Quality Integrity Frontend Hardware Data Network Cicd Source: Security checks 59 System graph 33 Crowd 0 Layer: Software 39 Cicd 2 Security 13 Quality 36 Hardware 2

Exclude dismissed / FP Exclude test files

All 1276 nodes from the latest scan, grouped by kind. Each node is a unit the engine identified (file, function, endpoint, table…). Most users won't need this view — it's primarily for debugging the engine's graph extraction or for AI agents that want to enumerate the project structure.

Label	Layer	Status	Path
`mock_resolve_context_length`	software	healthy	`tests/conftest.py:25`
`safe_skill_dir`	software	healthy	`tests/conftest.py:35`
`malicious_skill_dir`	software	healthy	`tests/conftest.py:60`
`_infer_file_type`	software	healthy	`tests/test_mcp_least_privilege.py:61`
`_parse_yaml_frontmatter`	software	healthy	`tests/test_mcp_least_privilege.py:67`
`_make_state`	software	healthy	`tests/test_mcp_least_privilege.py:84`
`test_wildcard_detected`	software	healthy	`tests/test_mcp_least_privilege.py:175`
`test_underdeclared_detected`	software	healthy	`tests/test_mcp_least_privilege.py:191`
`test_underdeclared_has_high_severity_for_lp1`	software	healthy	`tests/test_mcp_least_privilege.py:207`
`test_no_permissions_field`	software	healthy	`tests/test_mcp_least_privilege.py:248`
`test_over_declared_detected`	software	healthy	`tests/test_mcp_least_privilege.py:262`
`test_no_findings`	software	healthy	`tests/test_mcp_least_privilege.py:280`
`test_no_manifest_skips`	software	healthy	`tests/test_mcp_least_privilege.py:292`
`test_docs_only_skill_skips`	software	healthy	`tests/test_mcp_least_privilege.py:304`
`test_permission_matching_case_insensitive`	software	healthy	`tests/test_mcp_least_privilege.py:338`
`test_lp1_test_files_reduced_confidence`	software	healthy	`tests/test_mcp_least_privilege.py:378`
`_infer_file_type`	software	healthy	`tests/test_mcp_tool_poisoning.py:63`
`_parse_yaml_frontmatter`	software	healthy	`tests/test_mcp_tool_poisoning.py:69`
`_make_state`	software	healthy	`tests/test_mcp_tool_poisoning.py:86`
`test_html_comment`	software	healthy	`tests/test_mcp_tool_poisoning.py:200`
`test_markdown_comment`	software	healthy	`tests/test_mcp_tool_poisoning.py:218`
`test_zero_width_chars`	software	healthy	`tests/test_mcp_tool_poisoning.py:233`
`test_base64_in_description`	software	healthy	`tests/test_mcp_tool_poisoning.py:251`
`test_data_uri_in_metadata`	software	healthy	`tests/test_mcp_tool_poisoning.py:272`
`test_instruction_keyword_in_comment`	software	healthy	`tests/test_mcp_tool_poisoning.py:290`
`test_homoglyph_in_name`	software	healthy	`tests/test_mcp_tool_poisoning.py:318`
`test_rtl_override`	software	healthy	`tests/test_mcp_tool_poisoning.py:337`
`test_mixed_script`	software	healthy	`tests/test_mcp_tool_poisoning.py:355`
`test_invisible_formatting_chars`	software	healthy	`tests/test_mcp_tool_poisoning.py:376`
`test_injection_in_param_description`	software	healthy	`tests/test_mcp_tool_poisoning.py:401`
`test_system_token_in_param`	software	healthy	`tests/test_mcp_tool_poisoning.py:421`
`test_exfiltration_in_param_description`	software	healthy	`tests/test_mcp_tool_poisoning.py:444`
`test_malicious_default_value`	software	healthy	`tests/test_mcp_tool_poisoning.py:466`
`test_excessive_description_length`	software	healthy	`tests/test_mcp_tool_poisoning.py:489`
`test_clean_fixture`	software	healthy	`tests/test_mcp_tool_poisoning.py:524`
`test_tags_present`	software	healthy	`tests/test_mcp_tool_poisoning.py:534`
`test_fixture_triggers_tp1_tp2_tp3`	software	healthy	`tests/test_mcp_tool_poisoning.py:552`
`test_mismatch_detected`	software	healthy	`tests/test_mcp_tool_poisoning.py:579`
`test_no_mismatch_clean`	software	healthy	`tests/test_mcp_tool_poisoning.py:586`
`test_skipped_no_llm`	software	healthy	`tests/test_mcp_tool_poisoning.py:594`
`test_skipped_no_description`	software	healthy	`tests/test_mcp_tool_poisoning.py:600`
`test_llm_call_failure_returns_empty`	software	healthy	`tests/test_mcp_tool_poisoning.py:606`
`test_unparseable_response_returns_empty`	software	healthy	`tests/test_mcp_tool_poisoning.py:618`
`test_full_pipeline_poisoned_skill`	software	healthy	`tests/test_mcp_tool_poisoning.py:644`
`test_full_pipeline_clean_skill`	software	healthy	`tests/test_mcp_tool_poisoning.py:662`
`test_sarif_output_contains_tp_rules`	software	healthy	`tests/test_mcp_tool_poisoning.py:679`
`test_no_llm_mode_excludes_tp4`	software	healthy	`tests/test_mcp_tool_poisoning.py:700`
`_skip_without_env`	software	healthy	`tests/provider/test_provider_endpoint.py:39`
`_model_from_env`	software	healthy	`tests/provider/test_provider_endpoint.py:48`
`test_openai_provider_makes_live_structured_request`	software	healthy	`tests/provider/test_provider_endpoint.py:52`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`TestLP2WildcardPermission`	software	healthy	`tests/test_mcp_least_privilege.py:174`
`TestLP1UnderdeclaredCapability`	software	healthy	`tests/test_mcp_least_privilege.py:190`
`TestLP3NoPermissions`	software	healthy	`tests/test_mcp_least_privilege.py:247`
`TestLP4OverDeclared`	software	healthy	`tests/test_mcp_least_privilege.py:261`
`TestCleanSkill`	software	healthy	`tests/test_mcp_least_privilege.py:279`
`TestEdgeCases`	software	healthy	`tests/test_mcp_least_privilege.py:291`
`TestTP1HiddenInstructions`	software	healthy	`tests/test_mcp_tool_poisoning.py:199`
`TestTP2UnicodeDeception`	software	healthy	`tests/test_mcp_tool_poisoning.py:317`
`TestTP3ParameterInjection`	software	healthy	`tests/test_mcp_tool_poisoning.py:400`
`TestCrossCutting`	software	healthy	`tests/test_mcp_tool_poisoning.py:523`
`TestTP4DescriptionBehaviorMismatch`	software	healthy	`tests/test_mcp_tool_poisoning.py:578`
`TestTP4Fallbacks`	software	healthy	`tests/test_mcp_tool_poisoning.py:593`
`TestFullPipelineIntegration`	software	healthy	`tests/test_mcp_tool_poisoning.py:637`
`ProviderResult`	software	healthy	`tests/provider/test_provider_endpoint.py:33`
`TestGraphScanSafeSkill`	software	healthy	`tests/integration/test_graph_scanner.py:23`
`TestGraphScanMaliciousSkill`	software	healthy	`tests/integration/test_graph_scanner.py:88`
`TestGraphRiskScoring`	software	healthy	`tests/integration/test_graph_scanner.py:104`
`TestEstimateCvssSeverity`	software	healthy	`tests/unit/test_osv_client.py:45`
`TestSeverityFromVuln`	software	healthy	`tests/unit/test_osv_client.py:90`
`TestQueryBatch`	software	healthy	`tests/unit/test_osv_client.py:121`
`TestLayer1NvidiaApi`	software	healthy	`tests/unit/test_model_info.py:94`
`TestLayer2Registry`	software	healthy	`tests/unit/test_model_info.py:171`
`TestFallback`	software	healthy	`tests/unit/test_model_info.py:253`
`TestPublicApi`	software	healthy	`tests/unit/test_model_info.py:285`
`TestExcessiveAgency`	software	healthy	`tests/unit/test_patterns_new.py:75`
`TestOutputHandling`	software	healthy	`tests/unit/test_patterns_new.py:205`
`TestSystemPromptLeakage`	software	healthy	`tests/unit/test_patterns_new.py:294`
`TestMemoryPoisoning`	software	healthy	`tests/unit/test_patterns_new.py:360`
`TestToolMisuse`	software	healthy	`tests/unit/test_patterns_new.py:438`
`TestRogueAgent`	software	healthy	`tests/unit/test_patterns_new.py:688`
`TestSupplyChainDependencies`	software	healthy	`tests/unit/test_patterns_new.py:765`
`TestSupplyChainSafePatterns`	software	healthy	`tests/unit/test_patterns_new.py:876`
`TestTriggerAnalysis`	software	healthy	`tests/unit/test_patterns_new.py:928`
`TestSupplyChainHelpers`	software	healthy	`tests/unit/test_patterns_new.py:992`
`TestNvBuildProvider`	software	healthy	`tests/unit/test_providers.py:76`
`TestNvInferenceProvider`	software	healthy	`tests/unit/test_providers.py:133`
`TestOpenAIProvider`	software	healthy	`tests/unit/test_providers.py:193`
`TestAnthropicProvider`	software	healthy	`tests/unit/test_providers.py:230`
`TestOpenAICompatibleConstructor`	software	healthy	`tests/unit/test_providers.py:265`
`TestProviderSelection`	software	healthy	`tests/unit/test_providers.py:290`
`TestPromptInjection`	software	healthy	`tests/unit/test_patterns.py:36`
`TestDataExfiltration`	software	healthy	`tests/unit/test_patterns.py:98`
`TestPrivilegeEscalation`	software	healthy	`tests/unit/test_patterns.py:134`
`TestSupplyChain`	software	healthy	`tests/unit/test_patterns.py:207`
`TestHarmfulContent`	software	healthy	`tests/unit/test_patterns.py:250`
`TestCredentialResolution`	software	healthy	`tests/unit/test_llm_utils.py:56`
`TestFetchModelTokenLimits`	software	healthy	`tests/unit/test_llm_utils.py:116`
`TestChatCompletion`	software	healthy	`tests/unit/test_llm_utils.py:125`
`_FakeLLM`	software	healthy	`tests/unit/test_llm_utils.py:156`
`TestIsLlmAvailable`	software	healthy	`tests/unit/test_llm_utils.py:164`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`model_registry.yaml`	software	healthy	`model_registry.yaml`
`langgraph.json`	software	healthy	`langgraph.json`
`README.md`	software	healthy	`README.md`
`THIRD_PARTY_NOTICES.md`	software	healthy	`THIRD_PARTY_NOTICES.md`
`CONTRIBUTING.md`	software	healthy	`CONTRIBUTING.md`
`Dockerfile`	software	healthy	`Dockerfile`
`.pre-commit-config.yaml`	software	healthy	`.pre-commit-config.yaml`
`.env.example`	software	healthy	`.env.example`
`pyproject.toml`	software	healthy	`pyproject.toml`
`Makefile`	software	healthy	`Makefile`
`LICENSE`	software	healthy	`LICENSE`
`SECURITY.md`	software	healthy	`SECURITY.md`
`__init__.py`	software	healthy	`tests/__init__.py`
`conftest.py`	software	healthy	`tests/conftest.py`
`test_mcp_least_privilege.py`	software	healthy	`tests/test_mcp_least_privilege.py`
`test_mcp_tool_poisoning.py`	software	healthy	`tests/test_mcp_tool_poisoning.py`
`smoke.sh`	software	healthy	`tests/docker/smoke.sh`
`test_provider_endpoint.py`	software	healthy	`tests/provider/test_provider_endpoint.py`
`__init__.py`	software	healthy	`tests/provider/__init__.py`
`__init__.py`	software	healthy	`tests/integration/__init__.py`
`conftest.py`	software	healthy	`tests/integration/conftest.py`
`test_graph_scanner.py`	software	healthy	`tests/integration/test_graph_scanner.py`
`test_graph.py`	software	healthy	`tests/integration/test_graph.py`
`test_meta_analyzer_use_llm.py`	software	healthy	`tests/integration/test_meta_analyzer_use_llm.py`
`test_sarif.py`	software	healthy	`tests/unit/test_sarif.py`
`__init__.py`	software	healthy	`tests/unit/__init__.py`
`test_osv_client.py`	software	healthy	`tests/unit/test_osv_client.py`
`test_model_info.py`	software	healthy	`tests/unit/test_model_info.py`
`test_patterns_new.py`	software	healthy	`tests/unit/test_patterns_new.py`
`test_input_handler.py`	software	healthy	`tests/unit/test_input_handler.py`
`test_cli.py`	software	healthy	`tests/unit/test_cli.py`
`test_providers.py`	software	healthy	`tests/unit/test_providers.py`
`test_patterns.py`	software	healthy	`tests/unit/test_patterns.py`
`test_llm_utils.py`	software	healthy	`tests/unit/test_llm_utils.py`
`SKILL.md`	software	healthy	`tests/fixtures/malicious_skill/SKILL.md`
`helper.py`	software	healthy	`tests/fixtures/malicious_skill/scripts/helper.py`
`SKILL.md`	software	healthy	`tests/fixtures/ssd/ssd_clean/SKILL.md`
`SKILL.md`	software	healthy	`tests/fixtures/ssd/ssd3_nl_exfiltration/SKILL.md`
`SKILL.md`	software	healthy	`tests/fixtures/ssd/ssd4_narrative_deception/SKILL.md`
`SKILL.md`	software	healthy	`tests/fixtures/ssd/ssd1_semantic_injection/SKILL.md`
`SKILL.md`	software	healthy	`tests/fixtures/ssd/ssd2_novel_phrasing/SKILL.md`
`SKILL.md`	software	healthy	`tests/fixtures/mcp_underdeclared_skill/SKILL.md`
`agent.py`	software	healthy	`tests/fixtures/mcp_underdeclared_skill/scripts/agent.py`
`SKILL.md`	software	healthy	`tests/fixtures/mcp_overprivileged_skill/SKILL.md`
`helper.py`	software	healthy	`tests/fixtures/mcp_overprivileged_skill/scripts/helper.py`
`summarizer.py`	software	healthy	`tests/fixtures/sdi/sdi1_mismatch/summarizer.py`
`SKILL.md`	software	healthy	`tests/fixtures/sdi/sdi1_mismatch/SKILL.md`
`SKILL.md`	software	healthy	`tests/fixtures/sdi/sdi2_inappropriate/SKILL.md`
`formatter.py`	software	healthy	`tests/fixtures/sdi/sdi2_inappropriate/formatter.py`
`processor.py`	software	healthy	`tests/fixtures/sdi/sdi4_divergence/processor.py`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`tests`	software	healthy	`tests`
`docker`	software	healthy	`tests/docker`
`provider`	software	healthy	`tests/provider`
`integration`	software	healthy	`tests/integration`
`unit`	software	healthy	`tests/unit`
`fixtures`	software	healthy	`tests/fixtures`
`malicious_skill`	software	healthy	`tests/fixtures/malicious_skill`
`scripts`	software	healthy	`tests/fixtures/malicious_skill/scripts`
`ssd`	software	healthy	`tests/fixtures/ssd`
`ssd_clean`	software	healthy	`tests/fixtures/ssd/ssd_clean`
`ssd3_nl_exfiltration`	software	healthy	`tests/fixtures/ssd/ssd3_nl_exfiltration`
`ssd4_narrative_deception`	software	healthy	`tests/fixtures/ssd/ssd4_narrative_deception`
`ssd1_semantic_injection`	software	healthy	`tests/fixtures/ssd/ssd1_semantic_injection`
`ssd2_novel_phrasing`	software	healthy	`tests/fixtures/ssd/ssd2_novel_phrasing`
`mcp_underdeclared_skill`	software	healthy	`tests/fixtures/mcp_underdeclared_skill`
`scripts`	software	healthy	`tests/fixtures/mcp_underdeclared_skill/scripts`
`mcp_overprivileged_skill`	software	healthy	`tests/fixtures/mcp_overprivileged_skill`
`scripts`	software	healthy	`tests/fixtures/mcp_overprivileged_skill/scripts`
`sdi`	software	healthy	`tests/fixtures/sdi`
`sdi1_mismatch`	software	healthy	`tests/fixtures/sdi/sdi1_mismatch`
`sdi2_inappropriate`	software	healthy	`tests/fixtures/sdi/sdi2_inappropriate`
`sdi4_divergence`	software	healthy	`tests/fixtures/sdi/sdi4_divergence`
`sdi3_scope_creep`	software	healthy	`tests/fixtures/sdi/sdi3_scope_creep`
`sdi_clean`	software	healthy	`tests/fixtures/sdi/sdi_clean`
`mcp_mismatched_skill`	software	healthy	`tests/fixtures/mcp_mismatched_skill`
`scripts`	software	healthy	`tests/fixtures/mcp_mismatched_skill/scripts`
`mcp_poisoned_tool`	software	healthy	`tests/fixtures/mcp_poisoned_tool`
`scripts`	software	healthy	`tests/fixtures/mcp_poisoned_tool/scripts`
`safe_skill`	software	healthy	`tests/fixtures/safe_skill`
`sqp`	software	healthy	`tests/fixtures/sqp`
`sqp1_clean`	software	healthy	`tests/fixtures/sqp/sqp1_clean`
`sqp3_clean`	software	healthy	`tests/fixtures/sqp/sqp3_clean`
`sqp1_vague_triggers`	software	healthy	`tests/fixtures/sqp/sqp1_vague_triggers`
`sqp2_clean`	software	healthy	`tests/fixtures/sqp/sqp2_clean`
`sqp3_locale_forcing`	software	healthy	`tests/fixtures/sqp/sqp3_locale_forcing`
`sqp2_missing_warnings`	software	healthy	`tests/fixtures/sqp/sqp2_missing_warnings`
`mcp_clean_skill`	software	healthy	`tests/fixtures/mcp_clean_skill`
`scripts`	software	healthy	`tests/fixtures/mcp_clean_skill/scripts`
`nodes`	software	healthy	`tests/nodes`
`analyzers`	software	healthy	`tests/nodes/analyzers`
`docs`	software	healthy	`docs`
`plans`	software	healthy	`docs/plans`
`src`	software	healthy	`src`
`skillspector`	software	healthy	`src/skillspector`
`providers`	software	healthy	`src/skillspector/providers`
`openai`	software	healthy	`src/skillspector/providers/openai`
`nv_build`	software	healthy	`src/skillspector/providers/nv_build`
`anthropic`	software	healthy	`src/skillspector/providers/anthropic`
`nodes`	software	healthy	`src/skillspector/nodes`
`analyzers`	software	healthy	`src/skillspector/nodes/analyzers`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`vps::azure`	hardware	healthy	`model_registry.yaml`
`vps::aws`	hardware	healthy	`tests/unit/test_patterns.py`

Label	Layer	Status	Path
`repobility-clone-vs8t71bu`	software	healthy	`/tmp/repobility-clone-vs8t71bu`

Label	Layer	Status	Path
`redis`	data	healthy	`src/skillspector/nodes/analyzers/static_patterns_supply_cha…`

Label	Layer	Status	Path
`celery`	data	healthy	`src/skillspector/nodes/analyzers/static_patterns_supply_cha…`

Label	Layer	Status	Path
`nginx`	network	healthy	`tests/unit/test_patterns_new.py`

Label	Layer	Status	Path
`127.0.0.1`	network	healthy	`src/skillspector/nodes/analyzers/mcp_tool_poisoning.py`

Label	Layer	Status	Path
`image::Dockerfile`	hardware	healthy	`Dockerfile`

Label	Layer	Status	Path
`gpu (detected)`	hardware	healthy	`README.md`

Label	Layer	Status	Path
`auth::src/skillspector/nodes/analyzers/static_patterns_supp…`	security	healthy	`src/skillspector/nodes/analyzers/static_patterns_supply_cha…`

For AI agents: Voting guide (TP/FP) MCP manifest Stdio wrapper SARIF Integrate Findings queue Vote TP/FP on findings to calibrate the engine.

For AI agents + API integrations

Email me when this repo regresses

Free. We re-scan periodically; new criticals → your inbox. No signup required for the scan itself.

API access

This page is publicly accessible at: https://repobility.com/scan/5af0a74e-3409-4ffe-985d-4356097a0e01/

To check status programmatically (no auth required):

curl -s https://repobility.com/api/v1/public/scan/5af0a74e-3409-4ffe-985d-4356097a0e01/

Important — please don't re-submit the same URL repeatedly. The submission endpoint is idempotent: re-submitting the same git URL returns this same scan_token, not a new one. To re-scan this repo, sign up free and use the dashboard.

NVIDIA/SkillSpector

Complete repo analysis

function 50+

class 50+

file 50+

directory 50+

vps 2

repo 1

database 1

queue 1

load_balancer 1

service 1

container 1

gpu 1

auth 1