Plugin

harness-architect

harness-architect is a meta-tool that scans any target project and builds an end-to-end Claude Code harness — CLAUDE.md, settings, rules, agents, playbooks, hooks, and MCP — through a 9-phase orchestrated workflow. The plugin enforces a strict Agent-Playbook separation (WHO / HOW): agents define identity and hand off methodology to sub-agent-only playbooks, preventing the main session from bypassing guardrails. Every phase is reviewed by an independent red-team advisor that produces BLOCK / ASK / NOTE classifications, and cross-phase state is persisted to `docs/{request}/` so interrupted sessions resume deterministically. Target users: anyone — including non-developers — who wants a correct-from-day-one Claude Code setup on every new project without re-deriving the same 30-decision matrix each time. The plugin ships a 6-step "strict-coding" preset for complex coding projects and three Phase 1-2 branches (fresh setup / Cursor migration / harness audit).

• Published source
• Repository

Analysed

10 October 2026

Last observed

11 October 2026

What the code does

• Runs hooks automatically Hooks that execute without being called explicitly.

Having a capability is not wrongdoing; many tools need these to do their job. It tells you what to check before you install.

This is public metadata, not a safety certification or complete assessment. Static analysis cannot establish what a tool does at runtime.