Guards

Time left on the working turn, on the spinner line: from the task list's pace, else from how long your turns run. No model calls, no tokens. /eta turns it on or off.

See images, screenshots and HTML pages inline in the Claude Code chat

Play snake while Claude works. Every tool call Claude makes drops food into the game; the game pauses when Claude is done.

Catch wild creatures above the Claude Code prompt: where your prompt leads (water, grass, fire, bugs, plans, builds) decides what appears. Pick a partner that draws under the logo, beside the spinner and above the prompt, and levels up and evolves as Claude works. Needs function hooks (early access) and an interactive terminal.

Status bar above the prompt on the desktop surface: model, effort, folder, git branch, context usage, cost and elapsed time, with a hover map of the context window and a pane to remove built-in tools from it

Automated multi-agent code review cycle that spawns reviewers in parallel (plus Codex when available), applies fixes per CLAUDE.md policy, admits a commit only when a reviewer saw exactly what it records, and asks before a push, pull request, merge or release you did not request, unless you let it go further.

A card above the prompt: model, context and cache; the five-hour and seven-day limits with their pace; the running subagents with their type, model and effort, and the tokens they used; what the last turn sent and added, edits and errors.

The session's dashboard in one pane, opened by /dash: the model, the effort and presets that set both, ultracode and fast mode, each switched with one click; the context and quota gauges, the cost and the duration, /compact behind a confirmation, the session's name; the agents, commands, skills, plugins, MCP servers and hooks, a plugin or a server turned on or off from there; and the /config...

Bandeau projet (APK vs tel, git, quotas et conso en euros, tâches, sous-agents et build en pastilles) et garde-fous pour le dev Android avec Claude Code

Le petit Claude orange à droite du bandeau, qui réagit en silence à ce qui se passe dans la session

Context window details like the Desktop app's panel: /ctx opens a pane that docks beside the transcript, and a 📊 button above the prompt toggles it.

Progress bar above the prompt plus an agents panel (/agents-info) for any subagent work: running, finished and planned subagents with model, task progress, context, cost and time, each drawn as a crab whose costume follows its model and effort. Fork of johnnyvizz/claude-kit savvy-progress 1.2.0.

A pane that shows what Claude has done for you: tasks finished, an estimate of the hours that saved you, and what that time is worth. Tabs for Today, Week, Month and Year. /report opens it.

A side pane for the Home Assistant automation Claude is working on: triggers, conditions and actions as HA's editor nests them, what a save changed, which steps the latest run took, and a car-stereo dot-matrix display that acts it out. Works with @vortitron/home-assistant-mcp, hosted on Vome, linked through the Vome relay, or pointed at your own Home Assistant.

A Home Assistant dashboard in a side pane of Claude Code, working: live states, and controls that toggle, step and run as the dashboard's own do. Shows the dashboard Claude reads or edits, with what a save changed lit up, or one picked with /dash. Needs @vortitron/home-assistant-mcp.

A side pane for ESPHome in Claude Code: a map of the device from its YAML (board, network, buses, entities, pins, triggers) with what a save changed lit up, and the build Claude is running as it happens (phase, progress, the latest output, the result), over a dot-matrix chip on the bench. Needs @vortitron/home-assistant-mcp with esphome_activity.

A side pane for your Home Assistant's health: Vome's score out of 100 and what its check found, by severity with what to do, marked as Claude works on each, then a re-check rolls the score to the new one. Needs @vortitron/home-assistant-mcp with vome_health_report and the Vome integration.

Keeps secrets out of the transcript and out of the repo: redacts API keys, tokens, private keys and password URLs from tool results before Claude reads them, and refuses a Write, Edit or Bash command that embeds one.

Catches slop as Claude writes it: refuses a file edit that replaces code with a '... rest of the code' placeholder, and tracks TODO stubs, not-implemented throws, debug prints, @ts-ignore and other leftovers in a band above the prompt and a /slop pane.

Runs your test suite after every turn in which Claude changed source, and keeps the result above the prompt: green with the pass count, red with the first failing test. Detects bun, npm, pytest, cargo and go, or runs the command you set. Toasts on a regression and a recovery.

A podcast player inside Claude Code: subscribe to feeds, browse episodes in a pane, and play them with a now-playing bar above the prompt

The harness cockpit, a viewer: the phase every agent is in (PHASES panel, and main and the running agents in the band above the prompt), the advisor calls, what the insights learned (per-phase evidence, session outcomes from /insights, goals, the calibration and its proposals, with buttons to run /insights, apply, or open a calibration goal), Cortex memory health and grooming, the context...

Two rules no settings hook enforces, refused at the tool call: wiki pages and their generated ADR mirrors are written only through the Cortex wiki tool, and every git worktree lives at /.claude/worktrees/ / (one created there is registered with disk_hygiene.py).

Mods for the laa toolkit: blocks file changes and commits on the default branch, tracks the running laa pipeline above the prompt and in the footer, /laa-help, panes to triage learnings and review findings, and next-step suggestions.