Bugs

M1 — One agent end to end

Scheduled tasks A schedules table (project, cron, prompt or workflow) and a timer that creates the task the way the lead's create_task does, managed with lead tools and no UI, like workflows. Example: every Monday, run go vet ./... and gofmt -l and check docs/operations.md against the code. Claude Code Routines, Codex Automations, Copilot cloud automations, Devin, Factory and Conductor all schedule agents.

Answer a question again From an answered question in the Questions tab, give a different answer, which forks the task: a new task starts from the snapshot taken just before the original answer was delivered (the answer's snapshot_n), its agent continuing the conversation as it stood there (as Fork does from a snapshot), and gets the new answer. The original task carries on untouched.
Take both paths when unattended An agent working unattended (no human to ask) that hits a question with two plausible answers forks at it: the task carries on with one answer and a fork, started the same way as answering again, carries on with the other, so the human later compares both results instead of the agent guessing.

Phone app A native client for the phone over the same API, sharing the macOS client's SwiftUI code where it can. Every surveyed tool has one (Claude, Codex, Cursor and Conductor apps, GitHub Mobile, Superset for iPhone); it needs its own layout, since two half-width panes don't fit a phone.