ClaudeCodeMod

All shelves / Mods / Workflow

vbw

By swt-labs

vbw is a Claude Code workflow mod by swt-labs.

For VBW, which proves every requirement with a check, draws a Mission Control pane and a Comms strip, and denies risky commands during a run.

DrawsHolds a command

Install

The marketplace name is vbw-marketplace.

In Claude Code
/plugin marketplace add swt-labs/vibe-better-with-claude-code-vbw
/plugin install vbw@vbw-marketplace
Open the repo

It runs inside the Claude Code process. The side panel is only what the author's README names. Read the repo before you install, or run claude plugin validate on the plugin folder.

Version
2.1.0
License
MIT
Repo stars
86
Last change
2026-10-21

What it can reach

  • Draws in the session
  • Reads usage or context figures
  • Holds a command
  • Sees tool calls
  • Calls another program on this machine

Requires Claude Code 2.1.286 or newer, running Sonnet, Opus, or Fable, on macOS or Linux, or in WSL on Windows. It needs bash 3.2 or newer, jq, and git, in a git repository. Send the two commands as two separate inputs. The author says pasted together, Claude Code reads them as one command and the URL breaks. Restart Claude Code. VBW turns on Dynamic workflows and its status line the first time you use it, and /vbw:doctor checks the rest. /vbw:vibe is the next step, every time. The status line under the prompt shows the milestone, the plans done, the running step, the context, tokens, cache, cost, and lines changed. On Claude Code 2.1.287 and later, Mission Control is a side pane of where the project stands, in plain words, and /vbw-panel opens it. From 2.1.290, Comms is the strip above the prompt, with the helpers at work, or one line when VBW needs you. Each shows the running step's end time. A sound plays when a question needs you, on by default, and /vbw-sound switches it. /vbw-status, /vbw-why, and /vbw-todo answer at once, even mid-run, with no model turn. Guards check every command before it runs, block secret files, and deny writes outside the plan or the project during a run. While /vbw:loop is on, pushes, releases, publishes, and deploys are denied. A guard that breaks blocks the action it guards during a run. The author says VBW makes no network call. Workflow agents may also use Haiku.

The marketplace name is vbw-marketplace, and the marketplace source is ./plugin. VBW 2 is a rewrite of VBW 1. The Mission Control pane and the Comms strip are why this page files it under workflow. glanceflow draws a checklist above the prompt with the plan, the step Claude is on, and when Claude needs you. plan-progress draws live progress bars above the prompt, one row per task. This one tracks VBW's milestones and proofs, and its guards hold risky commands during a run. The guards are the plugin's settings hooks, and the pane and the strip are its hooks module. In the source, a Mission Control card runs the plugin's vbw kernel when you press it.

Demo

The VBW status line from its README: the milestone, 4 of 6 plans done, the running step until 14:20, then context, tokens, cache, cost, and the 5-hour and 7-day limits.
Screenshot of the author's GitHub page.

What it hooks into

  • session.startWhen a session starts
  • command.runWhen you run its slash command/vbw-panel/vbw-sound/vbw-status/vbw-why/vbw-todo
  • ui.closeWhen its pane closes
  • tool.callAround a tool call, where it can hold oneAskUserQuestionWorkflow
  • prompt.submitWhen you send a prompt
  • ui.renderDraws into the sessionPaneAbovePromptSpinnerPromptHintSessionModeInfoNoticeAskUserQuestionTurnDuration

Read from the on() calls in its hook modules. Open its source under Files to check what each one does.

Hooks 6

  • SessionStartbash "${CLAUDE_PLUGIN_ROOT}/hooks/session-start.sh"
  • Stopbash "${CLAUDE_PLUGIN_ROOT}/hooks/stop.sh" 2>/dev/null || true
  • PreToolUse · Bashulimit -t 5 2>/dev/null; r="${CLAUDE_PROJECT_DIR:-$PWD}/.vbw/record.json"; g=/dev/null; d="${CLAUDE_PROJECT_DIR:-$PWD}/.vbw/runtime/glance.json"; [ -f "$r" ] && [ -f "$d" ] && [ -r "$d" ] && [ "$d" -nt "$r" ] && g="$d"; o=0; [ -e "${CLAUDE_PROJECT_DIR:-$PWD}/.vbw/runtime/run-open" ] && o=1; jq -nc --rawfile glance "$g" --arg root "${CLAUDE_PROJECT_DIR:-$PWD}" --arg open "$o" -f "${CLAUDE_PLUGIN_ROOT}/hooks/guard-bash.jq" - "$r" "${CLAUDE_PLUGIN_ROOT}/hooks/end.json" 2>/dev/null || { [ "$o" = 1 ] && { echo "VBW guard: the guard failed while a run is open, so this call is blocked. Next: run vbw doctor to see why, then retry" >&2; exit 2; }; true; }
  • PreToolUse · AskUserQuestionjq -c -f "${CLAUDE_PLUGIN_ROOT}/hooks/approve-ask.jq" 2>/dev/null | { IFS= read -r in && bash "${CLAUDE_PLUGIN_ROOT}/hooks/approve-ask.sh" "$in"; } 2>/dev/null || true
  • PreToolUse · Read|Write|Edit|MultiEdit|NotebookEdit|Grepulimit -t 5 2>/dev/null; r="${CLAUDE_PROJECT_DIR:-$PWD}/.vbw/record.json"; g=/dev/null; d="${CLAUDE_PROJECT_DIR:-$PWD}/.vbw/runtime/glance.json"; [ -f "$r" ] && [ -f "$d" ] && [ -r "$d" ] && [ "$d" -nt "$r" ] && g="$d"; o=0; [ -e "${CLAUDE_PROJECT_DIR:-$PWD}/.vbw/runtime/run-open" ] && o=1; jq -nc --rawfile glance "$g" --arg root "${CLAUDE_PROJECT_DIR:-$PWD}" --arg open "$o" -f "${CLAUDE_PLUGIN_ROOT}/hooks/guard-file.jq" - "$r" "${CLAUDE_PLUGIN_ROOT}/hooks/end.json" 2>/dev/null || { [ "$o" = 1 ] && { echo "VBW guard: the guard failed while a run is open, so this call is blocked. Next: run vbw doctor to see why, then retry" >&2; exit 2; }; true; }
  • PostToolUse · AskUserQuestionout=$(jq -c -f "${CLAUDE_PLUGIN_ROOT}/hooks/approve-answer.jq" 2>/dev/null); [ -z "$out" ] || bash "${CLAUDE_PLUGIN_ROOT}/hooks/approve-answer.sh" "$out" 2>/dev/null || true

Files

The author's source, 15 files. They load when you open them.

Facts

Author
swt-labs
Category
Workflow
Marketplace
vbw-marketplace
Repo
swt-labs/vibe-better-with-claude-code-vbw
Folder
plugin
Version
2.1.0
License
MIT
Repo stars
86
Last change
2026-10-21
Language
Shell
Files
187

Other mods in Workflow

project-sentinelArjun0014Holds a commandDraws a status bar above the prompt for long runs, with milestones, context, cache, and the 5-hour window, a Control Room panel, and asks before risky calls.cortexesXBlueSkyFor Cortexes, a vault of your recorded sessions, draws its session-start menu as a band above the prompt, and offers a hand-off once the context is 70% full.proof9frankbriaHolds a commandTurns a bug into a CodeFRAME requirement and checks the ledger before each git commit Claude makes, warning by default or denying in block mode.directorEvolvingAgentsLabsHolds a commandKeeps a team's way of working as linked statements Claude plans from, and draws a status line and a band to run a waiting plan or accept a lesson it learned.bigquery-agent-analytics-modGoogleCloudPlatformStreams each session's prompts, model requests, tool calls, and subagents as rows to your BigQuery table, and says above the prompt when they land.next-stepsbradtraversyAfter each turn, has Haiku suggest two or three next prompts above the prompt, as buttons on the desktop that draft one, or 1, 2, and 3 in the terminal.critical-decisionsMucidaHas Claude record each call it made that you never asked for, with the reason, in a band above the prompt, red or yellow, with ok, rethink, and not critical.painelGabrielPauloTOpens a side panel, in Portuguese, of the running tool, Claude's latest thinking, its plan steps, the files it edited with line counts, subagents, and context.tavernweave-agent-skillsLiarMTTTDraws a Soul switch above the prompt that fills a persona's instruction into the prompt box, beside 23 skills for building SillyTavern character cards.hopper-hqFairMind-Gen-AI-StudioDraws a band above the prompt while another session waits for you or has finished, and opens hopper-hq, a browser office where each session has a desk.taskplannersmekaiOpens a board of the repo's .tasks/ Markdown tasks beside the transcript to move, add, or start one with Claude, and names the one in progress above the prompt.agent-taskskeshrathOpens agent-tasks' stage-gated pipeline in a /pipeline pane beside the conversation: tasks by stage, artifacts and comments, advance, send back, and complete.
All 396 in Workflow