sandbaseai/sandbase-harness

Local-first, self-hosted AI agent runtime and MCP bridge with sandboxed sessions, memory, credentials, audit/replay, and a local Console.

★ 500+ · sandbaseai/sandbase-harness source on GitHub · this plugin in the registry

sandbaseai/sandbase-harness is a DeepSeek Harness plugin rated C3 — powerful capability combined with sensitive behavior. It patches the dsh runtime, executes system commands, decodes base64 payloads, starts a network server.

Installable plugin — declares a dsh.bundle manifest

What it can do

Capability	Flag	Evidence
patches the dsh runtime	`runtime_patch`	./examples/deepseek-harness/cordis.yml
executes system commands	`exec`	×39 in authored code, e.g. scripts/prepare-install.mjs:2, scripts/prepare-install.mjs:2
decodes base64 payloads	`base64_decode`	×7 in authored code, e.g. src/api/routes/files.ts:163, src/api/routes/sessions.ts:279
starts a network server	`net_server`	×5 in authored code, e.g. scripts/smoke-release.mjs:153, scripts/smoke-release.mjs:156
reads credential-class env vars	`token_env`	MANAGED_AGENTS_API_KEY, MANAGED_AGENTS_ENVIRONMENT_KEY, MANAGED_AGENTS_SECRET_KEY, OPENAI_API_KEY

Outbound domains

api.minimax.io mcp.notion.com w3id.org api.minimaxi.com s3.amazonaws.com drivemcp.googleapis.com gmailmcp.googleapis.com calendarmcp.googleapis.com mcp.canva.com mcp.figma.com

Environment variables it reads

MANAGED_AGENTS_API_KEY OPENAI_BASE_URL OPENAI_API_KEY MANAGED_AGENTS_CORS_ORIGINS MANAGED_AGENTS_ENVIRONMENT_KEY MANAGED_AGENTS_URL MANAGED_AGENTS_HOME MANAGED_AGENTS_SECRET_KEY

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.