A collection of Claude Code plugins ("mods," requiring Claude Code 2.1.287+) that add live session instrumentation and guardrails. The flagship, mission-control, renders subagents, tool calls, and touched files in a side pane — blue for reads, orange for writes, green when done — via `/mission`. The Code view needs macOS, Chrome, and an image-capable terminal (Ghostty, kitty, iTerm2); the Who view is portable.
The rest is a practical toolkit: token-weather for context-fill and prompt-cache countdown, blast-radius to hold `rm -r`, force pushes, and migrations until confirmed, replay-theater for stepping through diffs, usage-meter for plan burn. Author openly marks the opinionated ones (merge-gate, rulebook-guard) as personal-setup forks to adapt, which is the right honesty. MIT licensed, CI badge present, install via the plugin marketplace with a no-install `--plugin-dir` trial mode.
The reels mod shows YouTube Shorts while Claude works and pauses when it's done, which is either the most honest UX decision of 2025 or a formal admission that watching the agent is a spectator sport. The snake mod settles it.
Worth asking: does mission-control's live rendering measurably change operator intervention rates, or is it a comfort blanket at four subagents and a hazard at forty?
reply