AIエージェント
agent-scan
Snyk's security scanner for AI agents, MCP servers and agent skills.
★ 3.1k スターApache-2.0
ガイドを見る →SkillSpectorは、NVIDIAが公開しているオープンソースのAIエージェントプロジェクトです（GitHub: NVIDIA/SkillSpector）。GitHubスターは19,702、フォークは1,722で、Apache-2.0ライセンスの下で公開されています。GitHubでの最終更新は2026年10月8日です。
NVIDIA's security scanner for AI agent skills - check before you install
Choose a skill directory, a single SKILL.md, a Git repository or a zip file to vet.
Install it as a uv tool, from a clone with make install, or use the Docker image.
Scan the target and review findings in the terminal, or export JSON, Markdown or SARIF.
Add the scan to your workflow or CI so unreviewed skills do not reach your agents.
Scans local skill directories, single SKILL.md files, Git repositories and zip archives.
Looks for prompt injection, data exfiltration, malicious patterns and supply-chain risks in skills.
Reports in the terminal or as JSON, Markdown or SARIF for CI/CD and IDE integration.
Run fully static with --no-llm, or enable LLM-assisted analysis by providing credentials.
A Docker build lets you scan without a Python environment.
An optional mcp extra packages SkillSpector for MCP-based workflows.
Get the source code from GitHub and enter the project directory.
git clone https://github.com/NVIDIA/SkillSpector.git
cd SkillSpectorThe recommended install registers skillspector as an isolated uv tool.
# Install uv first if needed, then install the tool from the repo - see README Quick StartPoint the scanner at a skill directory; add --no-llm for a fully offline static scan.
skillspector scan ./my-skill/ --no-llmIt is maintained by NVIDIA as an open-source project on GitHub.
Yes - SkillSpector is Apache-2.0 licensed and free to use.
No key is needed for static scanning (--no-llm). The optional LLM-assisted analysis uses a model provider you configure with your own credentials.
Compared with generic code scanners like Semgrep, SkillSpector is purpose-built for agent skills - prompt injection, exfiltration and skill supply-chain patterns that general SAST does not model.
Install it with uv per the README, then run skillspector scan on a skill directory and read the report.
Snyk's security scanner for AI agents, MCP servers and agent skills.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
Terminal security that stops malicious prompts, URLs, and payloads before they run.
YAO: rigorous engineering, evaluation, and governance for reusable agent skills.
Run chops locally: your ai agent skills, finally organized. a macos app to.
Voyager — the open-ended embodied Minecraft agent that writes its own skills and keeps learning for life.
LibreAIカタログに掲載されたAIエージェントプロジェクトの中で最もスターの多いもの — SkillSpectorが合わなかった場合に検討したい、現実的な候補リストです。
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
カタログの中でSkillSpectorとスター数が近いAIエージェントプロジェクト — 前後の棚もそのまま見て回れます。
Tongyi Deep Research - Alibaba's open-source deep research agent and 30B model
Devika - the open-source agentic software engineer that plans, browses and writes code
Kortix - the open-source AI operating system: agents, skills and memory as files in a repo
Open source agentic operating system
Give it a GitHub issue and it tries to fix the code for you - the original SWE-bench agent
Agent Zero AI framework