hashgraph-online/hol-guard
C2Open-source antivirus for AI agents: block risky tools, secret access, prompt injection, malicious packages, MCP servers, plugins, and skills at runtime.
★ 500+ · hashgraph-online/hol-guard source on GitHub · this plugin in the registry
hashgraph-online/hol-guard is a DeepSeek Harness plugin rated C2 — one powerful capability or sensitive behavior. It executes system commands, uses eval / new Function, decodes base64 payloads, installable-looking manifest in fixtures only.
A skill, not an installable plugin — ships SKILL.md with no plugin manifest
What it can do
| Capability | Flag | Evidence |
|---|---|---|
| executes system commands | exec | ×1 in authored code, e.g. scripts/dockerlabs/installed-guard-package-proof.mjs:106 |
| uses eval / new Function | eval | ×2 in authored code, e.g. dashboard/src/apps/app-catalog.ts:89, src/codex_plugin_scanner/guard/daemon/static/assets/chunks/app-catalog.js:48 |
| decodes base64 payloads | base64_decode | ×2 in authored code, e.g. dashboard/src/apps/app-catalog.ts:89, src/codex_plugin_scanner/guard/daemon/static/assets/chunks/app-catalog.js:48 |
| installable-looking manifest in fixtures only | dev_manifest | installable-looking manifest in fixtures only: tests/fixtures/deepseek-harness-good/package.json |
Outbound domains
hol.org twitter.com fb.me
Environment variables it reads
GUARD_INSTALLED_ORIGIN
How to read this
Levels measure capability surface and transparency, not maliciousness. A C3 plugin can be entirely legitimate — a desktop shell genuinely needs subprocesses. The point is that you can see this before installing. See the levels explained and how dsh plugins work.
Findings come from static analysis of shipped code; nothing is executed. Think a flag is wrong? Open an issue — every flag cites the file and line it came from.