Claude Code Safety and Permission Mods

Makes Claude ask before it touches your database migrations or runs destructive SQL, with the risk and what it found

Playable first-person voxel sandbox inside Claude Code

Context and usage meters, auto-resume, model router, agent cap, proof cards, context health, midnight and god modes, ship guard, Windows overlay

A Claude Code mod with switchable permission profiles: a safe baseline and named profiles you switch on and off

Choose the model each subagent runs on, by rule

Routes each subagent to a Claude model and effort that fits its task, judged by Jev

Run several Claude Code chats side by side safely: where the week's usage lands at this pace, a question before two chats edit the same file, heavy…

Animated desktop notifications when a session starts, a turn finishes, a tool fails, the context fills up, or Claude asks for permission

Claude Code for NVDA users on Windows: spoken announcements (finished, question, permission, background task), /time, /read and /quiet, a status line…

Pager: Claude Code pings you in Telegram when a task is done or it waits for your permission, and you text back the next prompt, /status, /last or…

在模型讀到之前遮罩台灣身分證／居留證號與有標籤的病歷號、健保卡號；擋下會把遮罩值寫回來源的工具呼叫

Replace protected people's names, Taiwan ID numbers, phone numbers and emails with placeholders before the model reads them

Claude Pets: a pixel pet game and retro dashboard for Claude Code (terminal, desktop app and VS Code), with plan usage, weather, your backlog story…

Stops Claude from killing, or running E2E tests against, another project's local dev server

Blocks em dashes in edits, shell commands and replies, and British spellings in prose files, commit messages and pull requests

Sample mod: denies edits to .env files

Required workflow checks with Claude Code function hooks

Asks before Claude runs destructive or slow-looking SQL through a DB CLI: DELETE, UPDATE without WHERE, DROP, TRUNCATE, full scans and cartesian…

Compacts before the 5-hour/weekly usage limit is hit, pauses until it resets, then resumes the task automatically

Roblox Studio safety layer for Claude Code (Studio MCP): RemoteEvent security audit, undo, Team Create protection, a confirmation before destructive…

rook by TestMu AI inside Claude Code: Claude runs rook against the agent it is building and gets failures back grouped into root-cause clusters with…

Pixel-art band above the prompt: folder, git and node; task progress with Clawd and his moods; a row of subagents; context, session and weekly…

Blocks dangerous shell commands (rm -rf, force push, hard reset, DROP TABLE)

Tu escudero en Claude Code: una sola fila que aparece solo cuando hace falta

A falling-sand sandbox in a side pane while Claude works

Explains each Claude Code sandbox block in plain words and turns recurring blocks into reviewed, one-press, undoable edits of your user settings

Context hygiene with safeguards: checkpoint each finished step, then reset the context to that save, never mid-project

SDD capability 04: semantic Analyze findings, read-only run guard (best effort), Analyze result state

Hides SecChain's secret values in prompts, tool results and instructions before the model reads them, through secchain mask

Keeps API keys, tokens and passwords out of the transcript, tool output, files, commits, PRs and issues

Bloqueia prompts e saídas de ferramentas que contenham chaves/segredos antes de chegarem ao modelo

Blocks git commit and git push from Claude when the change carries a secret file or a credential-shaped line, and flags secret files that are not…

Blocks API keys, tokens and private keys from being written to disk, put on a command line, or committed to git by Claude

Ask for credentials in a form, keep them in the macOS keychain, and run commands with them without showing the values to Claude

Two-way secret scrubbing: redacts credentials before the model sees them and blocks writing them into tracked files or shell commands

Collects API keys and other secrets in a private modal and writes them straight to their target file, redacting them from the session

Garde-fou : suspend les commandes Bash sensibles et demande Proceed / Cancel dans un panneau avant exécution

Keeps long sessions cheap and easy to pick up: warns before a session's cache goes cold and what coming back costs, nudges a wrap-up past a set size…

For repos where several AI agents (or you and an agent) work in the same folder: stops secret values from leaving your .env, pushes to public…

Blocks shell commands that wipe your home folder, the project, disks, protected branches or databases, and snapshots uncommitted work to a git ref…

Windows shell guard: fixes backslash drive paths in Bash, blocks quote-eating node -e / heredoc patterns, and catches PowerShell/Bash syntax sent to…

Spend Guard: live session cost band, a warning toast, and a hard stop that holds your next prompt once the session hits a dollar cap

Bloqueia escrita e DDL nos MCPs de SQL Server e comandos de shell que acessam banco; libera só após dois 'sim' do usuário

A departure board for gh stack: live stack pane, a one-line status above the prompt, and raw force-pushes, rebases and PR base edits on stacked…

Model, effort, permission mode, git, path, cost, and gradient bars for context and the 5H/7D limits above the prompt

Creator HUD above the prompt: live usage bars, a REC mode that masks secrets on screen, and an Explain pane that recaps the last turn

Keeps helper agents (subagents) from running up your usage: limit how many run at once, cap their model, and watch them work in a live pane

SwitchMod: useful extras for Claude Code, each one a switch