Real

/wazzap:real — "What's really up?"

Phase 2 of 4. Mode: Socratic, multi-turn. Probe one question at a time until the user can name the real problem under the surface wish. This is the framework's signature move — the part that earns the audit trail.

Goal

Produce real.md with a Q+A probing log and a clean problem statement that enables :approaches without rework.

Preconditions

1. Read parent SKILL.md for cross-cutting rules.
2. Resolve the active decision via <wazzap-root>/_active (or slug=<x> arg).
3. Read up.md to know the surface wish.
4. Soft-gate check: if no up.md exists, warn and write a stub up.md ("Skipped — picked up at :real on <date>"). Continue.

Steps

1. Open with the surface wish. Restate it back: "Surface wish from :up: '<wish>'. Now — what's really up?"

2. Probe one question at a time. Default budget: 3–5 questions. Hard cap: 8. After each round:

   • User answers.
   • Agent restates the real problem (current best draft) in one sentence.
   • Agent asks: "Does that land, or do we keep digging?"
   • User can stop early ("yes, that's it") or continue.

3. Probe categories — pick whichever the wish hasn't already covered. Don't run all of these; mix-and-match:

   Lens	Question pattern
   Symptom vs cause	"If you fixed the surface thing tomorrow, would the underlying frustration still be there?"
   Disappearance test	"If this problem vanished overnight, what would actually change in your day?"
   Who feels it	"Who's most paying the cost right now? You? Users? A teammate? Future-you?"
   Hidden constraint	"What can't you change, even if it would solve this?"
   Failure mode	"What does it look like 6 months from now if you do nothing?"
   Success signal	"How would you know this was solved without anyone telling you?"
   Trigger	"Why is this surfacing now, not 3 months ago or 3 months from now?"
   Reframe	"Is this a [decision / capability / risk / relationship / clarity] problem?"

4. Stop conditions. Stop probing when any of these is true:

   • User says "good", "that's it", "stop", or similar.
   • Agent has restated the real problem and the user agrees.
   • 8-question hard cap hit (write what you have, label remaining gaps in Unknowns).

5. Synthesize. Once probing is done, draft the structured sections (real problem, stakeholders, constraints, success criteria, non-goals, facts/assumptions/unknowns).

6. Show the synthesis to the user. Ask: "Does this land? Edit anything before I save?". Apply edits, then write real.md.

7. Update manifest. current_phase: real, append real to completed_phases, bump updated_at. Optionally update title if the framing shifted enough.

8. Suggest next command: /wazzap:approaches.

Probing technique notes

• One question per message. Never stack questions.
• Multiple choice when possible. Easier to answer than open-ended.
• Mirror the user's language. Use their nouns and verbs, not your reframe.
• Don't moralize. The Socratic move is to expose, not judge. If the user's "real problem" is "I don't want to have a hard conversation with my co-founder," capture that — don't soften it to "communication gap."
• Surface assumptions explicitly. When you hear "X always Y," ask "always? or just lately?"

real.md template

# Real problem (Phase 2)

**Slug:** `<slug>`
**Drilled on:** <ISO timestamp>

## Surface wish (from `:up`)

> <verbatim from up.md "The wish" section>

## Probing log

**Q1:** <agent question>  
**A1:** <user answer (verbatim or distilled with quotes around exact phrasing)>

**Q2:** <agent question>  
**A2:** <user answer>

<...continue for each Q+A...>

## The real problem

<One sentence. The thing that, if solved, makes the surface wish irrelevant or trivial.>

## Stakeholders

- **<role/name>** — <how they're involved, what they want>
- ...

## Constraints

- **Time:** <deadline or "open">
- **Money:** <budget bound or "n/a">
- **People:** <who's available, who's not>
- **Tech:** <hard tech limits>
- **Political/social:** <can't-touch topics, relationships at stake>
- **Other:** <whatever else came up>

## Success criteria (observable)

- <What you'd see/measure that proves it's solved>
- ...

## Non-goals

- <Explicit scope cuts — things the user said are NOT what they're trying to do>
- ...

## Facts / Assumptions / Unknowns

**Facts** (verified or self-evident):
- ...

**Assumptions** (treated as true, but not verified):
- ...

**Unknowns** (open questions that may or may not block `:approaches`):
- ...

Revisits (re-running :real later)

If real.md already exists, append, don't overwrite:

1. Read existing real.md.
2. Append a new section:

   ---
   ## Revisited <ISO timestamp>
   
   **Trigger:** <user-stated reason for coming back>
   
   ### Probing log (this round)
   ...
   
   ### Updated real problem
   <new one-sentence statement>
   
   ### What changed since last round
   - <e.g. "Stakeholder X was added", "Constraint Y was relaxed", "Surface symptom turned out to be different">
   

3. Update manifest:

   revisits:
     - phase: real
       at: <ISO timestamp>
       reason: "<trigger>"
   

4. Bump updated_at. Don't change completed_phases.

The original probing log stays untouched. That's the audit trail commitment.

Output to user (after writing)

✓ real.md written.

Real problem: <one-sentence statement>

Next: /wazzap:approaches to generate options.

Anti-patterns to avoid in :real

• ❌ Stacking questions ("Who's affected? Why now? What's the constraint?"). One at a time.
• ❌ Skipping the "current best draft" restatement after each round. That's how the user knows when to stop.
• ❌ Letting probing run past 8 questions. If you're at 8 and the real problem isn't crisp, write what you have and label the rest as Unknowns. Better imperfect than indefinite.
• ❌ Sanitizing the user's language. "I'm angry at my co-founder" is not "interpersonal alignment opportunity."
• ❌ Generating Q&A the user didn't actually answer. The probing log is verbatim or it's worthless.
• ❌ Overwriting an existing real.md. Always append on revisit.

Self-check before finishing

• Probing log has at least 3 verbatim Q+A pairs (or hit 8-cap with explanation).
• Real problem is one sentence, agreed by user.
• Stakeholders, constraints, success criteria, non-goals all present (use "Not yet identified" if genuinely empty, don't fabricate).
• Facts vs Assumptions vs Unknowns are labeled separately.
• Manifest updated: current_phase, completed_phases, updated_at.
• Suggested next command in output.