Guidance commands

Ask the agent where your project stands and what to do next — deterministic, free, and available on every project. · Recommended

VibeMap's agent can act like a product manager: point at what's thin, put the next steps in order, and run the one that matters most. You get that by asking for it — three commands in the agent panel — rather than by flipping a mode on.

Everything on this page is computed from your project's current state, so it costs zero tokens and works on every plan, including Free. The one paid, model-powered part of this section is the optional pre-flight interview, covered separately in Interviews and decisions.

Changed recently: guidance used to be a per-project Guide me / I'll drive toggle that pushed cards into the agent panel unprompted. That toggle is gone. The same analysis now arrives on request, through the commands below. Nothing was removed from what the agent can tell you — only the requirement to leave a mode switched on, and the cards that appeared in a panel most people kept closed.

The three commands

Type these in the agent panel's message box. They are ordinary slash commands, so /help lists them alongside the generation commands.

/guide — where you are and what's next

/guide answers in three parts, each omitted when there's nothing to say:

1. Next — the single recommended step, with the reason it's next and the exact command to run (for example, Run /gen-features). The recommendation is prerequisite-aware, so it's never a step you can't yet run.
2. Where you are — what's thin on the section you're currently looking at. On Features with none yet: "No features yet — they're the spine everything else hangs off." On Pages with orphans: "3 pages aren't tied to any user story."
3. The review plan — the top few outstanding items in priority order, each with the reason it matters.

Because it's pull-based, there's no throttling and no dismissal bookkeeping — asking twice in a row is a perfectly reasonable thing to do, and you'll get the current answer both times.

Actions are named as the command to type rather than rendered as buttons: now that /gen-features and friends exist, naming the command is the one-click affordance.

/next — do the recommended thing

/next runs whatever /guide would have recommended, without making you read the report first. It dispatches through exactly the same path as /gen-*, so your plan's tier gate, token billing, live progress, and the pre-generation Undo checkpoint all apply identically. /next chooses which generation to run; it never becomes a second way to run one.

If there's nothing outstanding, it says so and points you at /guide.

Check my coverage

A Check my coverage chip sits in the agent panel on every project route. It runs a deterministic gap report — features without stories, stories without criteria, pages without stories, personas without stories — and renders it as a card. Like the commands above, it's pull-based and free.

💡 Tip: Run Check my coverage right before Prepare for Dev. Closing gaps first makes the developer handoff far richer.

Asking for the agent's judgement

Separately from the commands, the conversational agent behaves differently when you ask for its opinion rather than issue an instruction ("why is there no login page?"):

• It proposes, instead of only advising. If acting on its own recommendation means changing the spec and it's confident in that change, it stages the change as a confirmation for you to approve or reject. Nothing is written until you approve. If it isn't confident, it asks exactly one question — the single question whose answer would most change the outcome — never an interrogation.
• It won't quietly reverse a decision. Once your decision log has entries, a request that contradicts one is surfaced as a clarification naming the decision it contradicts, rather than being silently planned.

What it costs

Nothing. /guide, /review, /next's recommendation step, and Check my coverage are all pure functions of data the panel already holds — no LLM call, no generation charge, no plan gate. /next then runs a real generation, which is billed exactly as if you had typed the /gen-* command yourself.

Only the optional pre-flight interview in Interviews and decisions calls a model, and it's opt-in per command.

⚡ Power-user hints

• /guide is route-aware: the "Where you are" section reflects the section page you're on, so running it from Pages and from Schema gives different middle sections.
• The commands /guide names submit the exact same task payload as the section-page Generate buttons — billing, live progress, and version control are identical.
• /review and /guide are the same underlying read. Use whichever word you'd naturally reach for.

↔ The traditional way

On a real team this is the product manager's job: noticing a feature has no stories, reminding everyone to write acceptance criteria before the handoff, and reviewing each artefact as it's produced. These commands encode that discipline into the editor, so a solo founder gets the same "someone's watching the plan" safety net without a second person in the room.

What's next

• Interviews and decisions — the optional --ask pre-flight interview and the decision log it builds.
• Prepare for Dev — consolidate the spec into a build-ready blueprint.