ztting01/dsh-agentenv-sandbox0

dsh-agentenv-sandbox

AgentENV execution-world bundle for DeepSeek Harness

AI Analysis

核心用途是为 DeepSeek Harness 提供 AgentENV 沙箱执行环境。适合需要安全、隔离的远程或本地沙箱环境来执行 Agent 代码的用户,需提前部署 AgentENV 服务。

Package
dsh-agentenv-sandbox
Version
0.1.4
License
MIT
Last updated
Aug 25, 2026

Install

$npx -p @deepseek-ai/dsh dsh plugin --profile web add github:ztting01/dsh-agentenv-sandbox

Configuration

Override the agentenv-runtime row in the profile's cordis.patch.yml. A later patch replaces the complete config, so restate every value you need.

KeyDefaultMeaning
apiUrlE2B_API_URL or http://127.0.0.1:8000AgentENV control plane
sandboxUrlE2B_SANDBOX_URL or apiUrlAgentENV data plane
apiKeyE2B_API_KEY, then AENV_API_KEYHost-only API credential
templateAENV_TEMPLATE_IDRequired template id or alias
sandboxIdomittedReconnect instead of create
cwdHarness process.cwd()Shared absolute remote cwd
timeoutMs3600000Rolling sandbox lease, renewed while Harness runs
securetrueSecure envd control traffic
onDisposekillkill or AgentENV pause
uploadWorkspacetrueInitial bounded host-to-sandbox upload
uploadMaxFiles50000Upload file-count bound
uploadMaxBytes512 MiBAggregate upload bound
uploadMaxFileBytes256 MiBPer-file upload bound
symlinkPolicycopy-internalCopy safe internal file/directory links, or use error/skip

To preserve a sandbox across Harness restarts, set onDispose: pause, then start with the recorded sandboxId. Automatic discovery and host write-back are intentionally not part of this MVP.