Add agent-handoff to Development & Code Tools
@WillowRyuchecks…add-agent-handoff → master1 files · +1 −0updated 1mo agoloading diff…
@WillowRyuchecks…add-agent-handoff → master1 files · +1 −0updated 1mo ago@WillowRyuchecks…add-agent-handoff → master1 files · +1 −0updated 1mo agoAdds agent-handoff — a Claude Code plugin (MIT) that splits coding work into a strict three-stage workflow with disk-backed state, so each stage can run in a fresh chat without losing context.
Bundle of four skills:
/setup-handoff — once per project; auto-scans manifests and writes .handoff/config.md (monorepo-aware: pnpm/turbo/lerna/nx/cargo/go workspaces)./plan — writes .handoff/plan.md with change list, parallelization groups, and verification plan./execute — applies the plan, tracks progress in .handoff/task.md. Forbidden from running tests/lint./verify — runs in a fresh chat. Reads only the plan and the diff (not the reasoning that produced the code) to catch missed requirements and out-of-scope changes.Agents working in a single context tend to rationalize their own bugs. The plan/execute/verify split forces independent verification, and the disk-backed state means each step survives a chat restart — important for longer features or context-window-heavy work.
Response language is the first interview question (any ISO code: en, ko, ja, zh, fr, ...). Every artifact is written in that language. Code, file paths, and command syntax stay native.
MIT. No telemetry — see PRIVACY.md.