xiaoxianxian/multi-proxy
C2多 Agent 编排中枢:把 Codex / Cursor / Hermes / DSH 等 AI 编程代理当员工统一管理——协议转换、成本台账、熔断、任务编排与 Web 面板。不是模型切换器,切的是活。
★ 0 · xiaoxianxian/multi-proxy source on GitHub · this plugin in the registry
xiaoxianxian/multi-proxy is a DeepSeek Harness plugin rated C2 — one powerful capability or sensitive behavior. It runs code at install time, executes system commands, decodes base64 payloads, starts a network server.
A skill, not an installable plugin — ships SKILL.md with no plugin manifest
What it can do
| Capability | Flag | Evidence |
|---|---|---|
| runs code at install time | install_script | postinstall: npm_config_build_from_source=true npm rebuild better-sqlite3 || true |
| executes system commands | exec | ×69 in authored code, e.g. free-llm/key-wizard.mjs:13, free-llm/key-wizard.mjs:13 |
| decodes base64 payloads | base64_decode | ×2 in authored code, e.g. workbuddy-shim/decrypt-token.js:58, workbuddy-shim/decrypt-token.js:91 |
| starts a network server | net_server | ×26 in authored code, e.g. claude-code-bridge/template-bridge.js:178, l2/llm-decomposer.demo.js:36 |
| reads credential-class env vars | token_env | AGNES_API_KEY_1, AGNES_API_KEY_2, ANTHROPIC_API_KEY, AUTH_LOCKOUT_MAX, AUTH_LOCKOUT_WINDOW_MS, AUTH_RATE_MAX |
Tool registrations
0 static tool registration point(s).
Outbound domains
api.deepseek.com api.openai.com api.moonshot.cn apihub.agnes-ai.com api.anthropic.com generativelanguage.googleapis.com openrouter.ai console.groq.com cloud.siliconflow.cn opencode.ai
Environment variables it reads
PROXY_AUTH_TOKEN MANAGER_AGENT_TOKEN MANAGER_URL PORT BRIDGE_MODE BINDING PROXY_L2_NOCANDIDATE MANAGER_PASSWORD JWT_SECRET PROXY_BUDGET_SNAPSHOT_FILE
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.