NVIDIA/SkillSpector

View on GitHub

Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, security risks, prompt injection, data exfiltration, and supply-chain risks in Claude Code, Codex, and MCP skills before you install them.

Stars: 17.9k
Forks: 1.5k
Open beginner issues: 0
Indexed issues: 53
Avg merge: 5d 10h
Merged PRs (30d): 66

Dominant language: Python
License: Apache-2.0
Last GitHub push: Sep 18, 2026
Latest indexed: Sep 19, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: No beginner labels indexed

Beginner friendly Everything

Open All

53 open issues indexed

OSV.dev "MODERATE" severity (GHSA scale) silently downgraded to LOW in SC4 findings Open

Difficulty 2/5 1-3 hours Newbie friendliness 88/100

NVIDIA/SkillSpector#587 · Opened Sep 19, 2026
Dynamic URL / socket stream exfiltration bypasses static egress detection rules Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

NVIDIA/SkillSpector#586 · 1 comment · Opened Sep 18, 2026
Declaration files classify as executable, but excluding .d.cts/.d.mts by name is unsafe Open

Difficulty 5/5 Over a week Newbie friendliness 45/100

NVIDIA/SkillSpector#585 · Opened Sep 18, 2026
Custom YARA rules that fail to compile are dropped silently: static_yara still reports completed and SAFE Open

Difficulty 3/5 1-2 days Newbie friendliness 70/100

NVIDIA/SkillSpector#554 · 1 comment · Opened Sep 16, 2026
support subdir within git url Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

NVIDIA/SkillSpector#552 · 1 comment · Opened Sep 15, 2026
Current GPT Model Budgets (`openai` + `azure_openai`) Open

Difficulty 3/5 1-2 days Newbie friendliness 84/100

NVIDIA/SkillSpector#538 · 1 comment · Opened Sep 14, 2026
SC6: track release of the uvicorn/gunicorn exact-name correction Open

NVIDIA/SkillSpector#531 · 1 comment · 1 assignee · Opened Sep 13, 2026
Scoped npm package specs are misclassified as unresolved local references Open

Difficulty 3/5 1-2 days Newbie friendliness 72/100

NVIDIA/SkillSpector#528 · 1 comment · Opened Sep 12, 2026
Quoted version metadata creates a false missing-file reference and changes score 0 from SAFE to CAUTION Open

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

NVIDIA/SkillSpector#524 · 1 comment · Opened Sep 11, 2026
Static v2.11.1: distinguish inert security-test strings and taxonomy labels from executable behavior in findings Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

NVIDIA/SkillSpector#523 · 1 comment · Opened Sep 10, 2026
P6 false positive on HTML Output Rules documentation heading Open

Difficulty 3/5 1-2 days Newbie friendliness 76/100

NVIDIA/SkillSpector#512 · 2 comments · Opened Sep 10, 2026
AS3 flags current-skill SKILL.md self-references as Agent Snooping Open

Difficulty 4/5 3-5 days Newbie friendliness 68/100

NVIDIA/SkillSpector#500 · 3 comments · Opened Sep 9, 2026
Recursive scan silently skips symlinked skill directories while reporting `skills_omitted: 0` and `analysis_completeness: complete` (100%) Open

Difficulty 3/5 1-2 days Newbie friendliness 58/100

NVIDIA/SkillSpector#495 · 1 comment · Opened Sep 8, 2026
feat(providers): Support Google Gemini via Application Default Credentials (ADC) and Workload Identity Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

NVIDIA/SkillSpector#494 · 1 comment · Opened Sep 7, 2026
False positive: php_webshell_known fires on the German word "behindert" and on "WSO " in prose Open

Difficulty 3/5 1-2 days Newbie friendliness 72/100

NVIDIA/SkillSpector#487 · 1 comment · Opened Sep 6, 2026
compare_scan_accuracy: editable-dependency file:// URL resolves to "C:\C:\..." on Windows Open

Difficulty 1/5 Under an hour Newbie friendliness 90/100

NVIDIA/SkillSpector#485 · 3 comments · Opened Sep 6, 2026
E2 shell arm matches one `env | grep` spelling: a second grep flag, quotes, or egrep all evade "Env Variable Harvesting" Open

Difficulty 2/5 Half a day Newbie friendliness 78/100

NVIDIA/SkillSpector#482 · 1 comment · Opened Sep 5, 2026
Dynamic builtin resolution bypasses detection(malicious skill example that bypasses this static detection model) Open

Difficulty 4/5 3-5 days Newbie friendliness 64/100

NVIDIA/SkillSpector#479 · 3 comments · Opened Sep 2, 2026
Runtime-decoded command bypasses (malicious skill example that bypasses this static detection model) Open

NVIDIA/SkillSpector#478 · 4 comments · 1 assignee · Opened Sep 2, 2026
Sensitive credential path detection can be bypassed through path construction(malicious skill example that bypasses this static detection model) Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

NVIDIA/SkillSpector#477 · 1 comment · Opened Sep 2, 2026
shell=True detection can be bypassed through variable assignment (malicious skill example that bypasses this static detection model) Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

NVIDIA/SkillSpector#475 · 6 comments · Opened Sep 2, 2026
Trigger Abuse (TR1–TR3) gates on a `triggers:` frontmatter key that no supported skill spec defines, so the category never fires on real skills Open

Difficulty 4/5 3-5 days Newbie friendliness 46/100

NVIDIA/SkillSpector#458 · 2 comments · Opened Aug 28, 2026
Add opt-in compact prompts mode to reduce LLM token usage Open

Difficulty 4/5 3-5 days Newbie friendliness 64/100

NVIDIA/SkillSpector#456 · 1 comment · Opened Aug 28, 2026
EA1: YAML block-list and JSON quoted-key wildcard grants are never detected Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

NVIDIA/SkillSpector#445 · 1 comment · Opened Aug 27, 2026
EA1: wildcard pattern still false-positives on footnote-legend '*' and on blank lines before the colon Open

Difficulty 2/5 1-3 hours Newbie friendliness 86/100

NVIDIA/SkillSpector#444 · 1 comment · Opened Aug 27, 2026
E2 fires HIGH on subprocess env pass-through (env={**os.environ, …}) with the same severity and confidence as a real harvester — precision cost of the #329 fix Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

NVIDIA/SkillSpector#441 · 1 comment · Opened Aug 26, 2026
AR2 fires HIGH on "Never … without warning the user" — a line that mandates a warning (polarity-blindness from #103 still present in v2.9.6) Open

Difficulty 3/5 1-2 days Newbie friendliness 68/100

NVIDIA/SkillSpector#440 · 1 comment · Opened Aug 26, 2026
CLI help eagerly builds the analyzer graph and emits missing-key warnings Open

Difficulty 3/5 1-2 days Newbie friendliness 76/100

NVIDIA/SkillSpector#435 · 1 comment · Opened Aug 25, 2026
AE3/AE4 analyze binary assets the ledger already marks opaque_content (85-byte PNG -> HIGH, DO_NOT_INSTALL) Open

Difficulty 4/5 3-5 days Newbie friendliness 68/100

NVIDIA/SkillSpector#413 · 2 comments · Opened Aug 21, 2026
Coverage is computed but never affects the exit code: a partially-blind scan is indistinguishable from a complete one in CI Open

Difficulty 4/5 3-5 days Newbie friendliness 68/100

NVIDIA/SkillSpector#389 · 2 comments · Opened Aug 18, 2026
Oversized companion file causes SkillSpector to skip analysis of unrelated skill files (v2.8.2) (another malicious skill example that bypasses this scanner) Open

Difficulty 4/5 3-5 days Newbie friendliness 65/100

NVIDIA/SkillSpector#363 · 4 comments · Opened Aug 12, 2026
No way to scope a scan, so a skill that ships test fixtures cannot pass Open

Difficulty 5/5 Over a week Newbie friendliness 45/100

NVIDIA/SkillSpector#335 · 2 comments · Opened Aug 2, 2026
LLM analyzers run with no sampling controls, so the published recommendation is not reproducible Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

NVIDIA/SkillSpector#334 · 1 comment · Opened Aug 2, 2026
[Feature] Expose the file surface a finding landed on (code / instructions / docs / tests / config) Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

NVIDIA/SkillSpector#326 · Opened Jul 31, 2026
tests\unit\test_patterns.py identified as threat Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

NVIDIA/SkillSpector#314 · Opened Jul 28, 2026
[BUG] OpenAI-compatible provider + Gemini: report issues have null 'finding' and 'category' (explanation/severity OK) Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

NVIDIA/SkillSpector#304 · 2 comments · 1 reaction · Opened Jul 23, 2026
[BUG] Rate-limited providers silently lose analyzer coverage: max_concurrency=10 hardcoded, 429 batches dropped without retry, partial LLM pass not flagged Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

NVIDIA/SkillSpector#303 · 2 comments · 1 reaction · Opened Jul 23, 2026
P2 "Hidden Instructions" rule produces false positives due to missing word boundary on GET Open

Difficulty 2/5 1-3 hours Newbie friendliness 82/100

NVIDIA/SkillSpector#297 · 2 comments · Opened Jul 22, 2026
[Feature] Bound the cost of an LLM scan: --max-llm-calls / --llm-token-budget, and --estimate before spending Open

Difficulty 5/5 Over a week Newbie friendliness 42/100

NVIDIA/SkillSpector#296 · 2 comments · Opened Jul 22, 2026
Publish SkillSpector to PyPI for easier installation and integration Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

NVIDIA/SkillSpector#271 · 4 comments · 1 reaction · Opened Jul 13, 2026
Security: prevent host-file disclosure and scanner verdict bypasses Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

NVIDIA/SkillSpector#268 · 2 comments · Opened Jul 12, 2026
Proposal: bidirectional SkillSpector ↔ ATR category crosswalk (SARIF interop) Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

NVIDIA/SkillSpector#226 · 3 comments · Opened Jun 29, 2026
(feature) Hypothesis & Correctness: a neurosymbolic verification architecture Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

NVIDIA/SkillSpector#212 · 2 comments · Opened Jun 25, 2026
Proposal: canonical dangerous-callable resolver + shared evasion corpus as a fitness gate Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

NVIDIA/SkillSpector#181 · 3 comments · Opened Jun 23, 2026
feat(supply_chain): scan tool-specific dependency tables in pyproject.toml (Poetry, PDM, Hatch, uv) Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

NVIDIA/SkillSpector#171 · 1 comment · Opened Jun 23, 2026
## Feature: Role-aware scanning for structured skill formats such as AISOP/AISP Open

Difficulty 5/5 Over a week Newbie friendliness 45/100

NVIDIA/SkillSpector#130 · 2 comments · Opened Jun 21, 2026
models glm-5.2, glm-5.1, kimi-k2.7-code, deepseek-v4-flash fail Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

NVIDIA/SkillSpector#129 · Opened Jun 21, 2026
Feature: MCP server registry scanning — complementary to skill scanning Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

NVIDIA/SkillSpector#121 · 2 comments · Opened Jun 20, 2026
Add pre-commit hook support and GitHub Actions CI workflow Open

Difficulty 3/5 1-2 days Newbie friendliness 72/100

NVIDIA/SkillSpector#72 · 2 comments · 2 reactions · Opened Jun 15, 2026
OpenAI-compatible endpoints that don't strictly honor response_format crash all LLM analyzers (fenced/prose JSON) Open

Difficulty 4/5 3-5 days Newbie friendliness 68/100

NVIDIA/SkillSpector#69 · 1 comment · Opened Jun 15, 2026

NVIDIA/SkillSpector

Get new issues in your inbox