Status line

A pane counting this repo's open GitHub issues, coloured by issue type and by who they are assigned to

奶蛙在终端里陪你写代码：像素头像、齁齁齁的台词、危险命令拦截，外加一个下班大笑彩蛋。

Verinoda in Claude Code: keeps the code index fresh while Claude edits, checks each edit for names that do not exist, reviews commits, and shows it all in a Turkish pane with an animated mascot.

A2A protocol client for Claude Code: Claude hands tasks to non-Claude worker agents over Agent2Agent and picks up their results.

Plan and track tasks inside Claude Code, kept as plain Markdown files in your project, with weekly sprints, a sprint board and agent teammates.

Claude assigns you the tasks only you can do. They wait in a My tasks pane, counted under the prompt, until you answer or reject them, so nothing Claude needs from you gets lost in the chat.

Keeps a long-running agent's working notes through every compaction: briefs the summarizer, re-attaches the notes after the summary, and shows them in a panel you can steer.

Shows how long the prompt cache stays warm, and when it expires, exactly how many tokens your next message will re-send uncached. No tokens spent.

Asks before Claude edits a file another chat on this machine changed in the last 30 minutes. Each chat keeps its own small ledger; no model calls, no network.

Trims long shell output before Claude reads it: keeps the head, the tail and every error line, saves the full output to a file Claude can open.

Stops Claude reading a file over 1,000 lines whole and gives it an outline tool (definitions with line numbers) so it reads only the range it needs. A retry still reads it whole.

Stops Claude re-reading a file it already read in this conversation when the file has not changed since; a deliberate second attempt still goes through.

Moves tools you have not used lately behind ToolSearch, so their descriptions stop riding along on every request; core tools and anything you used in your last five sessions stay loaded.

Asks before a single turn uses more than +5 session points or 500,000 uncached input tokens, and stops it cleanly if you say so. Checks between requests, never mid-command.

Picks the model and the reasoning effort for each task with TypeSafe's Jev, a System One decision model reached either through TypeSafe's own API or the Vercel AI Gateway. Routes subagent models at agent.spawn and the main loop's effort and model at turn.step, in both directions, falling back to the engine's built-in classifier when no key is set.

Watch the agent work: every step in plain English, with a live time-saved counter

Race Claude Code sessions on a task: live split-screen race track scoreboard

Failing tests spawn a pixel boss; every run that fixes tests lands a hit, zero failures is a KO with loot

A pixel pet that lives in a pane and reacts to what Claude does: eats on tool calls, gets sick on failures, panics on rm -rf, sleeps when idle, evolves as you ship

Gmail + Slack alerts inside Claude Code: toasts, status count, and an Alerts pane

Mission control for subagents and agent teams: who is running, what each is doing, who spawned whom, who is talking to whom, and how they overlap

Internet signal bars (latency to api.anthropic.com + periodic bandwidth) in the Claude Code status line.

Your Claude plan usage (5-hour and weekly limits), context fill and session cost, in the status line and as pills above the prompt.

把想清楚的开发任务派给本机 ZCode（GLM）执行，隔离在 worktree 里跑，红线加模型审批把关，用 git diff 和测试验收