Reskins

KOZMOS Aurora: northern lights above the prompt while the model thinks. Shifting curtains whose intensity and speed follow the effort level (low is gentle, max is wild), a calmer single wave while it responds, and a label like 'thinking · xhigh · 12s'. /aurora toggles it.

KOZMOS Compass: the single most useful hint under the prompt. In order: the 5-hour limit near its end with its reset, a context window near full, agents still working, uncommitted changes after edits, a draft left waiting; otherwise the engine's own hint. /compass toggles it.

KOZMOS Halo: a one-row mood line above the prompt whose color and motion tell what Claude is doing: calm blue breathing when idle, a violet flow while it requests and writes, a magenta shimmer while it thinks, the running tool's color racing across, a red flash when a tool fails, an amber blink when it waits for you. /halo toggles it.

KOZMOS Mantra: a themed spinner. The working line's word rotates through a pack of original gerunds (cosmic, zen, pirate, turkish, hacker) and its tail names the tool running now with its elapsed time, or the effort while the model thinks. /mantra switches, /mantra off hands the line back.

KOZMOS Stamp: a receipt on every finished turn. The closing 'Baked for 42s' line becomes a compact row of duration, tool calls by family as colored chips, tokens in and out, the turn's cost and the context fill with its change. /stamp toggles it.

An Improve button above the prompt (and /polish) rewrites your draft with Opus at low effort (/polish model picks another) using the prompt-master rules (MIT, nidhinjs/prompt-master); Undo puts the original back. Each press is one model call; its exact tokens are shown.

A side pane with robots: the main loop, every subagent and its tools, context, cost, rate limits and files touched

Shows the images you paste: thumbnails above the prompt, and larger pictures under each sent message in the chat; needs a kitty-graphics terminal such as Ghostty

Experimental kgai add-on for Claude Code with function hooks: checks each code-editing turn for a structural decision on the side, records it without a line in the transcript, and draws kgai's own tool calls as one line each. Requires the kgai plugin.

Splits each Claude Code turn's time into Claude, tools, you and other, and prints a /blame receipt for the session. Nothing leaves the machine. Claude Code 2.1.302+.

Part of AOT, the Agent Optimization Toolkit, a Mahal Systems offering. Design, build, measure and iterate the evals for a production agent: a repository contract (evals/), skills that walk commands C0-C18 with exit gates, a stdlib CLI that runs the agent k times and records every attempt, and a fixed report.json schema that Claude renders into a local HTML diagnostic.

Explains a shell command in one plain-English line, with its risk, right where you approve it, so you know what you're saying yes to.

To-do sidebar next to the chat: queue tasks, even while Claude is working. When Claude is clearly done, the to-do is checked off and the next one starts; otherwise the list stops with a reason and buttons (Continue keeps the order). Runs by default, never pauses on its own. Survives /clear. History per project. /todo, /todos help

RESCENE fan mod: WONI conducts while LIV, MINAMI, MAY and ZENA take the subagent and Orca worker tasks and report back, in Korean or English. 리센느 팬 모드: 원이가 지휘하고 리브·미나미·메이·제나가 작업을 맡아 보고한다.

Claude Code, with receipts. Plans with you, hands each task to a specialist agent, and keeps a session from calling work done until a verification is logged. Its hooks run in every session, a guard holds destructive shell commands for review, and setup can point your status line at OMCA.

Swear at Claude Code all you like: every prompt is cleaned before Claude reads it. Typos, translit, the wrong keyboard layout, swearing and CAPS go; the meaning, paths and numbers stay; your original rides along so nothing is lost.

Keeps long sessions going through API errors and rate limits: resubmits after a backoff when idle, waits out rate limits with a countdown, never loops, and stops the moment you type.

Keeps credentials out of your code, your shell commands and Claude's context: blocks hard-coded secrets in edits, asks before commands with literal keys, redacts secrets from tool results and flags them in prompts.

Prompt cache status band (c / ctx / rwc / sc / plan limits) with hover help, cold-send guard, keep-warm, /board, /handoff that also picks up handoff files, and an autonomous mode that clears and continues each handoff. Fork of Cache Keeper by Nate Herk.

Bookmarks and vim-style marks inside Claude Code conversations: mark a line, jump back, keep a reading position, browse your prompts, and let Claude cite earlier places as clickable links.

Kessoku Ultimate Bocchi Theme: a Bocchi the Rock!-inspired charcoal-and-pink skin for all of Claude Code (an engine theme plus every render site), a live Kessoku Band panel with the band on stage as animated pixel-art sprites under spotlights and their subagents as songs on a setlist (solo any member) and band subagent types Claude can delegate to by name, and themed Cache Tax, Blast Radius...

Keeps everything Claude writes in view and folds each stretch of tool calls and thinking between texts into one line you can expand

A side panel that keeps what a conversation leaves open: questions, your tasks, findings and unfinished asks