Desktop app that keeps a long project handoff-ready for any AI agent — tasks, decisions and a live handoff snapshot as diffable files. MCP hub. Tauri + Rust + React, local-first.
This repository also configures its own agents. See what agent-nextup tells them →
Skill Claude CodeCodex
A disciplined loop for bugs that resist a first look and for performance regressions — build a command that goes red on this exact bug, shrink it, test ranked hypotheses, then fix behind a regression test. Use when something is broken, throwing, flaky or slow and the cause is not obvious.
Skill Claude CodeCodex
Look back over a stretch of work and turn what went wrong into changes that steer the next session — automated checks first, then lessons, boundaries and pruned instructions, each backed by ledger evidence. Use in a Retrospect or Reflect phase, after a session that went sideways, or when the user asks what should…
Skill Claude CodeCodex
Write release notes or a pull-request body that a reviewer can act on — the smallest picture of the change, before-and-after evidence that it works, and whether it can be undone. Use in a Release phase for artifacts/release-notes.md, or whenever you write a PR description.
Skill Claude CodeCodex
Break agreed work into tasks the engine can schedule — vertical slices, each declaring what must land before it. Use right after a scope is settled, when a plan or spec needs turning into tasks, or when one task has grown too big for a session.
Skill Claude CodeCodex
Build behaviour test-first, one red–green slice at a time, at seams agreed before the first test. Use when implementing a coding task, when fixing a bug whose symptom can be reproduced, or when the user mentions TDD, red-green or test-first.
Skill Claude CodeCodex
Review finished work on two separate axes — the project's conventions, and what the task actually asked for — with one fresh reviewer per axis and the two reports kept apart. Use before a task is verified, in a Verify phase, or when asked to review a branch or a set of changes.
Skill Claude Code
Build every ready task in parallel — implementer sub-agents on the dependency frontier, each in its own worktree, merged onto one integration branch.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: