Our take
Activeregular commits and releases- Activity
- 78
- Community
- 86
- Issues
- 66
- Pull requests
- 53
+5,583 stars in the last 30 days
- Merges more pull requests than 96% of the projects we track
- More contributors than 78% of the projects we track
- Slower to merge pull requests than 89% of the projects we track
SkillSpector is an actively maintained, broad-contributor security scanner with useful install-gating outputs, but its findings are not a substitute for sandboxing or human review.
Good fit if
- You need to inspect AI agent skills before installation and want static checks with optional LLM analysis.
- You want machine-readable reports, CI exit codes, or an MCP tool to integrate scanning into an existing workflow.
- You can use Python 3.12 or newer within the supported range and accept the Apache-2.0 license.
Look elsewhere if
- You need runtime monitoring or containment; SkillSpector does not execute or sandbox scanned skills.
- Your review must remain fully local and you cannot use static-only scanning, since LLM analysis sends eligible file contents to the configured provider.
- You need comprehensive non-English or image-based attack detection, which the project lists as limitations.
All health signals
| Last commit | 2026-10-19 (today) |
|---|---|
| Commits, last 90 days | 100+ |
| Releases, last 12 months | 18 (latest v2.13.0, 2026-10-16) |
| Contributors | 100+ (top contributor: 35% of commits) |
| Issues closed, last 90 days | 23+ (typically closed in 5 days) |
| Pull requests merged, last 90 days | 95 (typically merged in 5 days) |
| Project age | 7 months |
Checked on 2026-10-20 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
25,724Forks
2,272Change since Nov 12, 2026: +231 stars·+23 forks
Overview
NVIDIA SkillSpector is a Python security scanner organized as a LangGraph workflow. It helps developers and security teams assess Claude Code, Codex, Gemini CLI, and MCP skills before installing them, looking for vulnerabilities, malicious patterns, prompt injection, data exfiltration, and supply-chain risks.
It accepts local files and directories as well as Git repositories, URLs, and zip archives. Static analysis runs without an LLM; optional semantic review can add context. SkillSpector reports findings and a risk recommendation, but it does not execute or sandbox the skill.
Key Features
- Scans skills from Git repositories, URLs, zip files, directories, or individual files.
- Detects risks across 17 categories, including prompt injection, unsafe code, data exfiltration, MCP permissions, and dependency vulnerabilities.
- Combines static checks, including Python AST analysis, with optional LLM semantic review.
- Queries OSV.dev for known dependency vulnerabilities, with a bundled fallback when the service is unavailable.
- Produces terminal, JSON, Markdown, and SARIF reports, with risk scores and recommendations.
- Supports finding baselines to suppress accepted issues and expose new findings in later scans.
- Offers an MCP server for agents that need to scan skills as part of an install workflow.
Use Cases
- Skill authors can check a skill’s scripts and instructions for risky patterns before publishing it.
- Security reviewers can triage third-party Claude Code, Codex, Gemini CLI, or MCP skills before approving them for internal use.
- Platform teams can add JSON or SARIF scans to CI and define whether findings or incomplete coverage should block a workflow.
- Agent-tool developers can use the MCP server to let compatible local agents request scans before installing a skill.
What you need
Detected in the repository
- Node.js (from package.json)
- Python >=3.12,<3.15 (from pyproject.toml)
- A Dockerfile, so it can run in a container
- A test suite and automated checks on GitHub Actions
License in plain words
Apache-2.0permissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the license, the NOTICE file and a note of your changes
- Share your changes: no
- Includes an explicit patent grant from the contributors:
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Install the CLI with uv and scan a skill:
uv tool install git+https://github.com/NVIDIA/skillspector.git
skillspector scan ./my-skill/
Use --no-llm for a static-only scan. See the README for provider setup, Docker use, MCP installation, and reporting options.
Alternatives
- apm: APM manages and installs agent resources with security and policy controls, while SkillSpector focuses on scanning skills for vulnerabilities and risky behavior.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| SkillSpector | Python | Apache-2.0 | 20.6k | Active |
| apm | Python | MIT | 3.9k | Active |
Considerations
- The project is active, with regular commits and releases and ongoing issue and pull-request handling. Its contributor base is broad rather than concentrated around one person.
- LLM review is optional, but when enabled the scanner sends analyzer-eligible file contents to the configured provider. Dependency names and versions are sent to OSV.dev for vulnerability lookups, including in static-only scans.
- Static analysis can produce false positives and does not assess runtime behavior. The README also notes limitations for non-English content, text embedded in images, and encrypted or binary code.
- Without access to OSV.dev, dependency checks use a smaller bundled fallback list. Scanning therefore has different coverage depending on network availability.
- The Apache-2.0 license applies to the project. Review the licenses of additional third-party software installed for your chosen setup.
Found this useful?
Share it with someone who would like SkillSpector.