Hi Aviv,
I forked this project and rewrote it from scratch as claude-protocol (published on npm as claude-protocol).
Submitted PRs to this repo went unreviewed. The underlying tools (beads CLI, Claude Code hooks API) changed significantly since the original release, so I continued independently.
What changed in v3
Solves open issues here:
Architecture:
- Removed 5 specialized agents (Scout, Detective, Architect, Scribe, Discovery) — duplicate built-in Claude Code capabilities
- Removed per-tech supervisor generation — 500+ lines of context with no benefit
- Removed agent personas — outdated prompting pattern
- 19 bash hooks → 8 Node.js hooks (cross-platform)
- Trigger-based rules instead of reference documents
- Checklist verification hook — blocks completion if requirements aren't met
- Session-start dashboard — surfaces open tasks, stale beads, recent knowledge
- Safe merge into existing projects — CLAUDE.md appended, settings.json merged, nothing overwritten
Install: npx claude-protocol init
Full rationale: docs/decisions-en.md
Open invitation
If you return to active development, happy to collaborate. Your original concept is the foundation — credit preserved in README and LICENSE.
Hi Aviv,
I forked this project and rewrote it from scratch as claude-protocol (published on npm as
claude-protocol).Submitted PRs to this repo went unreviewed. The underlying tools (beads CLI, Claude Code hooks API) changed significantly since the original release, so I continued independently.
What changed in v3
Solves open issues here:
Task(). Relates to EnterPlanMode not blocked — Claude hijacks orchestration with native plan mode #4 (cc @simon-77).Architecture:
Install:
npx claude-protocol initFull rationale: docs/decisions-en.md
Open invitation
If you return to active development, happy to collaborate. Your original concept is the foundation — credit preserved in README and LICENSE.