v1.0

First public release.

One expensive model orchestrates, five pinned cheaper subagents do the work.

• core/: global CLAUDE.md, five agent files (scout, researcher, builder, refuter, debugger), /task-session command
• extras/: /brief, /task-status, /task-close, per-repo CLAUDE.md template, settings.json, bucket templates
• validate_kit.py: 88 structural checks

See README.md for install steps.