pkgscan.dev · PKGSCAN by Yeeth Security
Scan it before you install it.
pkgscan.dev is a free, multi-engine scanner for software packages. Paste a package name, upload an archive, or look up a SHA-256 to see whether any of several independent security engines flags it as malicious or reports known vulnerabilities. Every result is public — no account is needed to read one.
What you can scan
- npm packages —
pkg:npm/{name}@{version}. Install scripts, obfuscated payloads, credential access, and known CVEs. - PyPI packages —
pkg:pypi/{name}@{version}. Wheel and sdist analysis for install-time code and exfiltration patterns. - VS Code and Open VSX extensions —
pkg:vscode-extension/{publisher}/{name}@{version}. Extension code, bundled dependencies, and publisher identity. - Cargo, Go, Maven, NuGet, and RubyGems packages — the same report shape. Paste a PURL for any other package the search bar does not list.
- AI agent skills — Smithery, SkillsMP, ClawHub, skills.sh, Guild.ai, AgentPM, and GitHub. Checks include prompt injection, data exfiltration, privilege escalation, and MCP tool poisoning.
- Any file by SHA-256 — reputation lookup without uploading the artifact.
How detection works
Each submission runs every engine that is eligible for the artifact. Engines report separately, so you can see which one found what, and the summary keeps malware findings separate from vulnerability findings.
| Engine | Class | What it contributes |
|---|---|---|
| Argus | Malware | Yeeth Security's content-analysis engine, built from live campaigns and run on Open VSX publish traffic. Scans any artifact, including skill files |
| SkillSpector | Malware | Agent-skill analysis: prompt injection, exfiltration, tool poisoning |
| GuardDog | Heuristic | Static rules for install hooks, obfuscation, credential access, and exfiltration |
| OSV, Grype, deps.dev | Vulnerability | Known CVEs and published advisories matched to the package identity |
A clean scan is evidence, not proof. Static analysis cannot see runtime payloads, encrypted second stages, or native binaries. Treat a clean result as one signal among several, or compare Argus with the other engines.
How a scan runs
- Submit. Drop a package archive, pick a registry package, paste a skill PURL, or enter a SHA-256.
- Fan out. The Eclipse SCINTX gateway routes the artifact to every engine that accepts that input.
- Engines vote. Each engine reports its own verdict and evidence; the summary aggregates them.
- You decide. Open any engine's detail and decide whether it breaks your threat model.
Plans and pricing
Free accounts get 5 scans a month with the full findings from every engine. Paid plans add volume, API keys, and team seats. Quotas reset on the UTC calendar month.
| Plan | Price | Scans / month | API access |
|---|---|---|---|
| Free | $0 | 5 | — |
| Hunter | $19/mo | 100 | Included |
| Operator | $99/mo | 500 | Included |
| Team | $199/mo | 1,000 | Included + team seats |
| Enterprise | Custom | Custom | Included + SSO, dedicated support |
See pricing and scan limits for what counts as a scan and how API keys work.
Documentation
- How to check an npm package for malware before installing it
- How to check a PyPI package for malicious code
- How to check a VS Code or Open VSX extension for malware
- How pkgscan detects malicious packages
- Pricing and scan limits
- Privacy and scan data
- Compare Argus with GuardDog, OSV, Grype, deps.dev and SkillSpector
Developers and AI agents
The same scan engine is available from scripts and agents. Paid plans can create API keys in Settings and call the HTTP API with Authorization: Bearer pkg_…; the app also exposes a Model Context Protocol server so an agent can call scan_purl and get a full verdict inline. Setup and endpoint reference live in the developer docs.
Common questions
Is pkgscan.dev free?
Yes. Reading any report is free and does not require an account. Creating an account gives you 5 scans a month. Paid plans raise the monthly allowance.
Does pkgscan store the code I upload?
Scan artifacts and results are stored so the report stays available at a stable URL, and results are public by design. Do not upload private source code, credentials, or personal data. See privacy and scan data.
Which ecosystems are covered?
npm, PyPI, VS Code Marketplace and Open VSX extensions, Cargo, Go, Maven, NuGet, RubyGems, and AI agent skills. Detection depth varies by ecosystem because engine eligibility depends on the artifact and the registry data available.
How accurate is it?
Every engine is published with its findings, and heuristic hits are kept separate from malware verdicts. The comparison pages show what each engine reads, covers and contributes.