Toasts

Mask token-like strings in tool output before they reach the conversation

Verbatim Jev-guided compaction for Claude Code sessions.

A desktop mascot overlay that reacts to turns, tool calls and subagents

Holds a prompt back once your 5-hour or weekly usage is at or over a limit you set (90% by default); sending it again goes through. No tokens spent.

Tells you when a newer version of an installed claude-pro-kit mod is out, and /kit-update installs it. One small download per session, no tokens spent.

Records every skill use with the skillmon CLI

/compact-jev: Jev-guided verbatim compaction for Claude Code, called through Vercel AI Gateway. Never replaces /compact or auto-compaction.

Clawd as a desktop pet for Claude Code: floats above your windows, shows Running / Needs input / Ready / Blocked, click to jump back into the session, quick chat and stop from the pet.

各ターンの所要時間とトークン消費をトーストで表示する

Lossless compaction for Claude Code: nothing is summarized; everything removed is archived verbatim with provenance and restorable by id (Jev or a local ruleset). /lossless to inspect and restore.

Switches the Claude in Chrome extension between named browser profiles with select_browser, which needs no approval click, and picks the default before the first browser call of a session.

Routes heavy Bash commands (test suites, builds, Docker) to a second machine over ssh, with the PATH export and working directory already right, and the commit synced first.

Bouncer for Claude Code: enforce condition-based markdown guardrail rules on tool calls — deny or annotate any call whose content matches, instead of file-path-only rules.

Jev-scored context pruning for Claude Code through OpenRouter: at compaction every tool call and result is scored by TypeSafe's Jev decision model, stale ones are dropped or truncated, everything kept stays verbatim. An optional gate truncates oversized tool outputs as they arrive.

Verbatim Jev-guided compaction for Claude Code and Codex sessions.

Every tool call your agent makes, checked before it runs. Verifies each pending tool call against the session's plan using TypeSafe AI's Jev model, then allows it, asks you, or blocks it.

Go make coffee: a desktop notification, a chime, and optionally a phone push when a long Claude turn finishes or Claude is waiting on you.

Reusable prompts one command away: /snip review drops your saved review prompt into the prompt box, with {{selection}}, {{branch}}, {{date}} and {{args}} filled in.

A seatbelt for YOLO mode: blocks catastrophic shell commands (rm -rf ~, mkfs, dd to a disk, fork bombs) and asks before risky ones (force-push, reset --hard, DROP TABLE, curl | sh), even with permissions bypassed.

Keeps Claude's file edits inside the project: refuses writes outside it (symlinks resolved), never lets a tool touch .git internals, and asks before it changes lockfiles, CI workflows, migrations or secrets.

Defuses prompt injection in tool output: strips invisible Unicode that hides text from you but not from the model, and flags web pages, files and command output that carry instructions aimed at Claude.

Stops data exfiltration and surprise network calls: checks every curl, wget, ssh, scp, git push and WebFetch against allow and deny lists, and asks before data goes to a host you haven't approved.

Keeps secrets out of Claude's context: API keys, tokens, private keys and passwords in tool output are replaced with [REDACTED:kind] before the model or the transcript ever sees them.

Stops Claude from installing hallucinated or typosquatted packages: every npm, pip, uv, poetry, cargo and gem install is checked against its registry first. Packages that don't exist are blocked; brand-new, barely used or lookalike packages ask you first.