How I tested the routing mod: every task ran twice, default and routed. A blind
How I tested the routing mod: every task ran twice, default and routed. A blind Opus judge with the repo open got both answers as X and Y in random order and checked every path, line number and config value. Full results, all three rounds: https://benenewton.com/blog/jev-agent-routing-claude-code?utm_source=threads&utm_medium=social&utm_campaign=every-claude-code-subagent-inherits-the-main-session-s-model-14
How I tested the routing mod: every task ran twice, default and routed. A blind Opus judge with the repo open got both answers as X and Y in random order and checked every path, line number and config value.
Full results, all three rounds: https://benenewton.com/blog/jev-agent-routing-claude-code?utm_source=threads&utm_medium=social&utm_campaign=every-claude-code-subagent-inherits-the-main-session-s-model-14
See the original source for installation instructions.
How I tested the routing mod: every task ran twice, default and routed. A blind Opus judge with the repo open got both answers as X and Y in random order and checked every path, line number and config value.
Full results, all three rounds: https://benenewton.com/blog/jev-agent-routing-claude-code?utm_source=threads&utm_medium=social&utm_campaign=every-claude-code-subagent-inherits-the-main-session-s-model-14
Domain-Driven Design scaffolding plugin for Claude Code with a function-hook guard that blocks secrets in DDD memory writes.
universal-modder is an MIT-licensed Claude Code plugin that can identify game engines, read real code, build mods, generate assets with fal, test in-game, and produce showcase videos.
Gives Claude a register_assumption tool and shows the turn's assumptions, decisions and considered-but-not-done items above the prompt