Guards

After a compaction, adds one message with exact facts from before it: your latest prompt in full, up to 4 earlier ones, the files edited, the todo list and the last failed shell command. At most 4,000 characters. No model calls.

Lists skills you have not used lately in this project by name only, so their descriptions stop riding along on every request; the Skill tool and /name still load any of them. No model calls.

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.

Shows you, above the prompt, the critical decisions the agent made without being asked: red to rethink, yellow to know, each with a rethink button. Works out of the box for any kind of project; /critical-areas auto lets the agent learn what is critical in yours.

The best all-in-one mod pack for Claude Code: a usage limits and context HUD, a guard with undo for rm -rf and git reset --hard, .env and secret protection, and a receipt for every turn.

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

Push guard: before a git push to a public GitHub repo (or making one public) it scans what would go out for Supabase project IDs, keys, tokens, emails, local paths, private repo names and .env files, and asks in a one-key popup. /pushor toggles. Hungarian or English.

Stops Claude reading .env and key files, masks API keys in command output, and lets you hand Claude a key through a pane without it ever seeing the value

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+.

Opinionated pipeline from a feature, bug, or refactor description to a verified, publish-ready implementation: scoped interview, wave-based plan, parallelized subagent execution, cross-model branch review, on-device UI verification via claude-in-chrome, and GitHub reconcile — layered on the superpowers skill set, with a doctor/learn loop that benchmarks and improves the pipeline from session...

Session pane: context, cost and turns, the prompt cache with the time until it expires, keep warm and a guard that holds a prompt that would write an expired cache again, the last turn's tokens, cache, tools and refused calls, a task list Claude keeps with a switch that keeps Claude working while tasks remain open, an autopick switch that answers a question left unanswered with its recommended...

Test-first workflow for coding agents, without the spec ceremony. A PreToolUse hook keeps source edits behind a mirror test and protects existing assertions; the flow drives one vertical slice at a time, red to green, through the project's own gate. Ships its own tdd and grill skills; includes the Jev layer (model and effort routing, compaction, review) and the interface (usage row, gate...

See the decisions Claude made in each turn, why it made them, and steer future turns with keep and avoid rules.

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.

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.

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

Quality-first routing: picks main-loop effort (medium to max) per request, assigns model and effort to subagents by kind and weight, checks Workflow agent() staffing, records judgments locally; the top-model guard is opt-in

Speaks Claude Code turn results aloud with VOICEVOX on Windows, macOS and Linux (CLI); long turns are summarized by Haiku inside the session (no extra API key, no session context touched). Voices: VOICEVOX:四国めたん, VOICEVOX:ずんだもん

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.

Consultation and prompt engine for Claude Mods. The module is the core; [probe. ] tables in probes.toml are the consultants and [prompt. ] tables author a system-prompt section, a tool description or a command description. The only carrier is the mod: a CLAUDE_*_CARRIER handle set to mod or left empty/unset arms the probe here; any other value loudly refuses the calls this probe acts on (the...

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.

Keeps a live local page for a long Claude Code task: what needs you, what was built, what was verified, and what was decided without you. Hooks record it without spending model tokens, the page follows light, dark or your system's setting, and when the task ends the same page is the report.

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.