debug

"${CLAUDE_PLUGIN_ROOT}/bin/vbw" config models 2>&1 || true
"${CLAUDE_PLUGIN_ROOT}/bin/vbw" config effort 2>&1 || true
"${CLAUDE_PLUGIN_ROOT}/bin/vbw" workflows on 2>&1 || true
"${CLAUDE_PLUGIN_ROOT}/bin/vbw" interview 2>&1 || true

Write to the user's level, explanation depth and involvement (above; "-": plain words), and pass them to the workflow as "profile": {"level", "depth", "involvement"} (none for "-").

The problem, in the user's words: $ARGUMENTS

If that is empty, ask what is wrong (what they did, what happened, what they expected). VBW needs the Workflow tool: the workflows line above turns it on (say so in one line when it did); without it, say why and stop. (Outside a VBW project the models line shows an error; omit models.)

Start the Workflow vbw:investigating with args {"problem": "<the problem>", "models": <the JSON above>, "effort": <the effort JSON above>, "profile": ...}. When it returns, tell the user the root cause, its evidence, rejected hypotheses, confidence and the proposed fix, then ask whether to fix it now. If yes, start the Workflow again with args {"problem": "<the problem>", "fix": <the diagnosis it returned>, "models": ..., "effort": ..., "profile": ...}: one Debugger fixes the root cause, adds a regression test, commits and verifies (vbw prove in a VBW project). In a VBW project, then record that the user asked for the fix, naming each file it returned by its path: vbw decide "Debug fix: <root cause>, in <file>, <file>" "the user asked for the fix (/vbw:debug)", so QA takes the change as the user's, not as a deviation from a plan. Report what changed.