plan-generation

AdvPlan-Bench is an offline benchmark for adversarial evaluation of structured plan-generation agents, comparing plans with BLUE-vs-RED advantage and diagnostic metrics across 150 synthetic scenarios to study response-budget sensitivity and multi-agent critique.

Adeptly is a free, open-source tool that teaches users how to use more of Claude Code's features by baking them into project plans. Version 0.5 adds an automated pipeline of roles (Architect, Builder, Medic, etc.) that executes workflows locally, with dry-run mode and a safety flag to prevent accidental live runs.