Symptom

_step_memory_dir (and every other _step_* in init_cmd.py) hard-codes its step number via step_header(N, title). That number is correct for the --advanced flow (where memory-dir is step 3 of 10) but misleading in the two preset flows where memory-dir is the first interactive step the user actually sees:

First-time users reading 3. without seeing steps 1 and 2 get a nagging "did I miss something?" signal, and the surrounding log ordering looks bizarre. (This may have contributed to the confusion in #371 — 3. implies there was a previous step to return to.)

Root cause

packages/memtomem/src/memtomem/cli/init_cmd.py — each _step_* function opens with step_header(<hardcoded_int>, "<title>"). The integer is a property of the --advanced 10-step sequence, not of the function's position in whatever run_steps list actually invokes it. packages/memtomem/src/memtomem/cli/wizard.py:step_header takes the number as a parameter but has no way to know the step's real position inside the current run_steps queue.

Suggested directions

Pick one:

1. Pass position through run_steps. run_steps knows i and len(steps); it could stash state["_wizard_position"] = (i + 1, len(steps)) before each step call. Step functions read that instead of hard-coding. One-time wiring, no signature change to step functions.
2. Make step_header take a title only and derive the number from a state["_wizard_position"] seed. Same idea, different seam — the number lives in the helper instead of the step.
3. Drop the number entirely from the preset flows (show only the title), or use word markers like "First. …" / "Next. …". Less structured but zero risk of drift.

Direction 1 or 2 is the right choice if we want the full sequence numbered consistently. Direction 3 is the lowest-touch escape hatch.

Acceptance criteria

• In all three flows (advanced, --preset X, default-interactive preset), step headers reflect the actual position within the flow the user is taking.
• No step function has to know or care which flow it runs in.
• Existing step tests continue to pass.
• New test that exercises the number display for at least one step in each flow (can use CliRunner().invoke(init, input=...) and grep the output).

Related

• mm init: "b" does not work at step 3 (Memory Directory) #371 — "b" not working at step 3 (user expected a step-2 to return to; my PR fix(init): "b" at Memory Directory navigates back to preset picker (#371) #418 fixed the navigation but left the number display as-is).
• fix(init): "b" at Memory Directory navigates back to preset picker (#371) #418 — combined preset-picker + follow-ups into one run_steps. Step-number display was explicitly deferred from that PR's scope.