Prompt

Session modes that are enforced, not requested: investigate, draft, build, ship. Denies the tool calls a stance rules out, auto-switches on phrases like "no code", and shows a pixel-art class badge above the prompt.

Rules that are enforced, not remembered: deny, ask, rewrite or note on any tool call, subagents included. A pixel-art trap for every mistake you already wrote down.

An animated pixel-art Claude mascot above your prompt, with a theme of the day picked by Claude itself / Une mascotte Claude animée en pixel art au-dessus du prompt, avec un thème du jour choisi par Claude

Saves every conversation to a per-project SQLite memory and lets Claude search it

Picks the model and reasoning effort for each task with TypeSafe's Jev. Routes subagent models at agent.spawn and subagent and main-loop effort at turn.step, with optional main-loop model routing. Falls back to the engine's built-in classifier when no key is set.

Picks the Claude model and reasoning effort for each turn with Cloudflare Clef, a fast decision model, on Workers AI or fully local (llama-server, Ollama).

Shows images pasted into the prompt as thumbnails above it

Macrodata Refinement while your agents work: find the scary numbers and bin them.

Live cricket commentary for Claude Code: every tool call is a ball, with a scoreboard, a pixel big screen, crowd sounds and an Indian commentator who speaks.

Keeps the prompt you last typed on a row above the prompt line.

Makes Claude explain its work in the fastest format to understand: STE-style text, diagrams, throwaway HTML pages or explainer videos, alone or combined, chosen by project and context.

Shows the track playing in Apple Music or Spotify above the prompt, with cover art, play/pause, skip, volume, shuffle and repeat controls, a share-link copy and an Up next pane.

/issue N loads a GitHub issue into the prompt; /issue alone opens a pane of open issues

Jev Semaphore for Claude Code: Jev / Kev model routing for subagents, with a live allocation pane, agent graph and context-exchange band

Skills and guard hooks: ai-code-cleanup, interrogate, shape-task, verify-change, setup, a shared-worktree git guard, and a guard against reading API-key config files

Cross-model review and delegation: /mm:review and /mm:run, built on mmrun

A per-project progress board: after each typed turn that edited files or made a commit, the plugin records its work as nodes through a Sonnet completion, kept under ~/.claude/progress/ or committed with the project; the plugin mirrors it to an optional Artifact canvas

Inline image previews in the terminal (Ghostty, kitty): under every image Claude reads, above the prompt for pasted images, and on demand with /img

Talk to your Claude Code sessions from other sessions, your voice, or any local program. Meetings between two sessions, and a talk button that reads replies aloud.

Usage limits above the prompt, a pane with background work and the file each open Claude Code session is on

Shared Switchboard gateway, setup, status and reversible Claude startup integration.

Developer dashboard pane: what needs you, every Claude session and subagent, usage limits, git and PRs

Paste a link with a short question and Claude's answer opens with a verdict for your projects: use it, later or skip, the effort, the cost and one line why

Every prompt carries your local date, weekday and the week ranges, and a reply whose dates miss the period you asked about gets flagged