Security for Multi-platform

CLI/CI scanner — not Aqua Platform management UI/RBAC/policy SaaS. Vs Grype/Snyk: Aqua OSS default with broad target surface; Aqua Inc monetizes a separate commercial platform. You still triage and patch.

Signing CLI + public-good Sigstore infra — not a full policy admission controller (pair with Kyverno/OPA). Offline/air-gap needs different PKI. Registry must support referrers/OCI artifacts cleanly.

Maintainer marks feature-complete (security patches; focus shifting to Betterleaks). Official Action for GitHub org repos needs a free license key from gitleaks.io — personal accounts exempt. Not a full ASVS/SAST suite. False positives need allowlists.

Vuln scanner — not Trivy's IaC/secrets swiss-army, not Cosign signing, not Anchore Enterprise governance UI. You still triage CVEs. Anchore monetizes Enterprise separately.

PoW slows mass scrapers; determined headless solvers still pass. Can impede Archive/RSS unless you allowlist. Not Cloudflare WAF. Distinct from Traefik/Caddy proxies.

Vs Trivy/Grype (broader/image-first scanners) and Syft (*SBOM producer*): OSV-Scanner is the OSV.dev-native vuln CLI, not a replacement for every Trivy filesystem/misconfig check and not the SBOM generator. Complements those tools; do not expect identical CVE sets. Networked API unless you go `--offline`.

GPL-3.0 copyleft — same class as ShellCheck; not MIT. Vs Docker (the engine), dive (*layer/size forensics*), and lazydocker (Docker/Compose *TUI*): Hadolint is the *Dockerfile linter*, not a runtime, not layer diffs, not a container dashboard.

Vs Promptfoo and DeepEval: AI-Infra-Guard is an AI-infrastructure red-team platform, not only an eval runner; vs generic SAST, it understands agent/MCP/skill attack surfaces. Preserve the mandatory attribution — “Based on Tencent Zhuque Lab AI-Infra-Guard” — with the repository link in your deployment and derivative notices. Some tools also require an LLM_API_KEY, so a self-hosted scanner can still carry provider cost and credential risk.

Vs Agent Substrate and coding sandboxes: LitterBox is a malware/payload detection sandbox with an MCP surface, not a general coding workspace or agent runtime isolation layer. The README says “Development use only”; production isolation is an unsolved risk. Windows administration and Linux KVM/Windows-container requirements can be substantial, and GPL-3.0 copyleft still applies to distributed modifications.

Vs SkillSpector / AI-Infra-Guard / destructive-command-guard: SkillSpector gates agent skills; AI-Infra-Guard is a broader red-team platform; dcg blocks destructive host commands — MCP Scanner’s catch is scanning MCP servers/tools/prompts/resources. Full Cisco AI Defense / VirusTotal paths need keys and network; treat OSS engines as the free bar and cloud engines as optional cost/trust.

Vs SkillSpector / mcp-scanner / destructive-command-guard / AI-Infra-Guard: SkillSpector scans skills pre-install; MCP Scanner audits MCP servers; dcg is a host deny-hook; AI-Infra-Guard is red-team platform — HOL Guard’s catch is runtime local policy/antivirus across agent actions including MCP calls. Policy false positives can stall agents; optional cloud is not required — keep local-first if that is the point.

Linux only (kernel 5.6+ recommended). Not every syscall is implemented — exotic workloads fail or fall back. Heavy file and network I/O pay a measurable overhead versus runc. This is a runtime, not an SDK or an orchestration product: you still wire Docker/K8s/agent harnesses yourself. Vs Firecracker/Cloud Hypervisor (microVMs) and plain seccomp: gVisor is the userspace kernel middle path.

Enforces at admission time, so a bad policy (or an unavailable webhook with fail-closed settings) can block deploys cluster-wide — roll out in audit mode first. YAML policies get awkward for complex logic; Rego-based engines express that more cleanly. Vs kubescape (posture and compliance scanning) and topaz (application authorization): Kyverno is the in-cluster admission policy lane.

A heavy Java desktop app — the 12.1 release needs a 64-bit JDK 21, and Python 3 for the debugger and PyGhidra. The UI is dense and the learning curve is steep. Check that you are allowed to reverse-engineer the target before you start.