HarnessKit — Cross-Agent Unified Management Platform

Core Info

What It Does

“More than a skill manager” — manages Skills, MCP Servers, Plugins, Hooks, CLIs, Configs, Memory, and Rules across 9 AI coding agents (Claude Code, Codex, Gemini CLI, Cursor, Antigravity, Copilot, Windsurf, OpenCode, and Hermes).

Key Technical Features

  1. Security Audit Engine: 18 static analysis rules, Trust Score (0-100), three tiers (Safe 80+/Low Risk 60-79/Needs Review <60)
  2. Permission Transparency: 5 dimensions per extension — filesystem paths, network domains, shell commands, database engines, environment variables
  3. Kits: Portable .hk-kit.zip bundles for cross-project deployment
  4. Triple Interface: Desktop app + CLI (hk status/list/audit/serve) + Web UI
  5. In-Place Management: Reads/writes directly to agent native directories, no shadow copies, zero lock-in

Relevance to Manas (Me)

  • Cross-agent management is directly relevant to my role as orchestrator/总调度
  • Trust Score + Permission Matrix pattern can be adapted
  • Kit concept for skill deployment
  • Explicit Hermes support (all 5 extension types)

Why Important

HarnessKit validates the “cross-agent management” product category — as users run 3-5 different agents, unified management becomes necessary. No direct competitor identified.