PluginNeeds setup

sandbase-harness

sandbaseai64377Updated today

Persistent managed-agent runtime for DSH via a native bundle and stdio MCP, with sandboxed sessions, audit, and replay.

agent-harness agent-observability agent-plugin agent-runtime agent-sandbox ai-agents ai-infrastructure audit-logging deepseek-harness devops docker dsh-plugin harness-engineering kubernetes

Practical score

87 / 100

Setup

Needs setup

License

Apache-2.0

Language

TypeScript

Install

Quick start

Copy the command, or hand the whole task to your harness agent.

git clone --branch v0.3.8 --depth 1 https://github.com/sandbaseai/sandbase-harness.git

Setup required

Install with your agent

Please install the DeepSeek Harness plugin "sandbase-harness" for me. - GitHub repository: sandbaseai/sandbase-harness (https://github.com/sandbaseai/sandbase-harness) - Type: a Cordis plugin (installed into a DSH profile) - This plugin may need extra configur…

From the README

A cleaned excerpt from the project's own documentation.

SandBase Harness English 中文 AI-readable project meta

Open on GitHub

Project details

Repository: sandbaseai/sandbase-harness
by: sandbaseai
License: Apache-2.0
stars: 643
Forks: 77
Updated: Nov 11, 2026
Created: Jul 11, 2026
Language: TypeScript
Indexed from: awesome-0xsline · topic

Visit homepage

Related entries

Similar tools from the same corner of the ecosystem.

View all

Skill

deepseek-harness-docker

runzhliu

Community Docker and Kubernetes packaging for DeepSeek Harness (@deepseek-ai/dsh), with a hardened image, Compose stack, Helm chart, Web UI, and headless CLI.

agent-runtimeai-agentscli

164Needs setup

83/100

Plugin

dsh-jev-plugin

luobosibing2

Opt-in Jev/Luna decision layer for DeepSeek Harness: skill and file selection, task supervision, tool-output filtering, approval assistance and recorded-stage navigation; features are disabled by default.

agent-harnessai-agentscordis

207Needs setup

83/100

Plugin

dsh-maze

lamost423

Execution maze for agent sessions: main path, detours and backtracks on one timeline, per-step data tracks (tools, tokens, context pressure), deterministic analysis (outcome evidence, behavior signals) and multi-session comparison.

agent-observabilityagent-trajectoryai-agents

90Ready to use

77/100