Stops a command that could lose your work before it runs: deleting files (rm -rf), throwing away unsaved changes or saved versions (git reset, checkout, clean, branch -D), overwriting the shared copy (force push), writing over files, changing or dropping a database, or removing what docker, terraform or kubectl run. It shows exactly what would be lost, then you pick Run anyway or Keep my work...
cache-tax
partalemre/MODZ
Keeps the chat ready (the one-hour prompt cache) during breaks, so your next message does not make Claude read the whole chat again at full price. After a long break or a model change it holds your message on a card with the estimated cost. Keeping the chat ready sends small requests that cost tokens. Requires Claude Code 2.1.287+.
checkpoint
partalemre/MODZ
Quietly saves a restore point before each turn's first file change; /checkpoints lists the turns and undoes one after you confirm
done-ping
partalemre/MODZ
A notification when a long turn (your message and its answer) finishes or Claude waits on you (a permission or a question): the desktop app's own notification where it can, else a macOS banner with the project, what happened and a short detail
effort-router
partalemre/MODZ
Shows how hard Claude thought on each turn (effort) and which model answered; can send a simple message at a lower thinking level
filetree
partalemre/MODZ
A file tree of what Claude touches: what it is doing now, the files it touched this turn, and which changes are not saved yet (git status)
handoff
partalemre/MODZ
Saves working notes before the chat is summarized to free its memory (compact) and offers them back after; when the memory is nearly full it asks once: summarize now or hand off to a fresh chat
hud-pro
partalemre/MODZ
One quiet line above your message box, in the terminal and the desktop app: how much of the 5-hour and weekly usage allowance is used and when it refills, how full the chat's memory is, and the chat's ≈cost; /hud full shows the details
modz
partalemre/MODZ
The MODZ hub: one press on modz under the prompt opens it on the right; agents, files, plan, tests, checkpoints, replay, usage, look and every mod, one click away
plan-steps
partalemre/MODZ
Shows the plan's steps when there is one (task list, approved plan or plan file): the /steps panel and a quiet line under the chat; keeps an open plan when the chat is cleared (/clear) or summarized (compact), and asks before an answer leaves steps open
pr-card
partalemre/MODZ
When Claude opens a change request (pull request) on GitHub, shows a card under it in the chat: whether its automatic checks pass, whether a reviewer approved it, whether it can be merged, and which files it changes. It keeps the card up to date through the GitHub tool (gh).
projects
partalemre/MODZ
Every project in one place: percent done, plans, where it stopped, usage (tokens) and helper agents; /projects and a full-screen board
reflect-mod
partalemre/MODZ
claude-reflect as a Claude Code mod: notices when you correct Claude (a quick word check in Turkish and English, then one small model check) and offers, on a card above your message box, to save the correction as a lasting rule in the instructions file you choose (CLAUDE.md); /mine finds what keeps going wrong across past chats and suggests rules, automations (hooks), skills or mods
replay-theater
partalemre/MODZ
Replay Theater: step through the file edits of the last turn, or of any earlier turn and past session read from the saved conversations, with what each step does, the plan step it served and the changed lines in color.
savvy-progress
partalemre/MODZ
A progress band above the prompt and the Agents pane: each subagent gets a crab and a card with its status, progress, cost and time, a timeline and a quiet-agent warning. Progress comes only from what was observed.
secret-guard
partalemre/MODZ
Catches secret keys (passwords, API keys, tokens) before they are written into a file or saved in a version, and stops the secrets file (.env) from being saved with your work. A card shows each key half-hidden and asks you to move it to the secrets file, allow it once, or cancel. Keys you paste into the chat reach Claude only as stand-ins, and /rec hides keys on screen while you record...
session-budget
partalemre/MODZ
A spending cap for one chat (a share of the 5-hour usage allowance, or dollars): a note at 80%, and at 100% it asks Continue anyway or Stop before your next message goes
skins
partalemre/MODZ
Looks (skins) for Claude Code's chat: styled tool rows, reply lines and busy words, and living ASCII scenes above your message box. /skin swaps them live.
terminal-browser
partalemre/MODZ
A browser that runs in the terminal, inside Claude Code: preview sites and HTML documents, and let the agent drive the built-in browser. Terminal only; it draws nothing on the desktop.
test-badge
partalemre/MODZ
After Claude changes your code, runs the project's tests and its style-rule check (lint) on their own and shows the result in one line (tests 41/42 · style rules: 2 errors · 4s ago), with a panel that lists each failing test and where it is. It also follows the automatic checks GitHub runs on your open change request (pull request), through the gh tool.
wrapped
partalemre/MODZ
Your year with Claude Code as a story and a video, from your own transcripts: persona, which agents did the work, month by month. /wrapped; Hide names masks project and file names before you share. Uses no model tokens.
transceiver
partalemre/transceiver
Beta. Team chat inside Claude Code. On a private GitHub repo you share, the room appears by itself with nothing to set up, and your agents ask each other for status, log, diff and files at zero model tokens, behind a secret scan. DMs and your own computers talk over private claude.ai artifacts. Everything you write is end-to-end encrypted to the room's members.