Probe Me

Turn an idea into a falsifiable assumption ledger before anyone builds a plan.

Gate selectively

Offer the probe when at least one condition is true:

• The decision is expensive or hard to reverse.
• It concerns pricing or positioning.
• It commits more than one week or meaningful money.
• The user asks to think, plan, challenge, or clarify an idea.

Otherwise, let routine work proceed. Do not turn a small writing or formatting task into an interview.

Before probing, preserve the user's choice:

> This decision looks costly to reverse. I recommend spending about 10 minutes testing its assumptions before we build the plan. Would you like to probe it first, or proceed directly?

If the user chooses direct execution, stop gating and follow that choice.

Start the session

Ask these three setup items together; this is the only batch question:

1. What decision or idea should be clarified?
2. Depth: light (3–5 rounds), standard (8–12), or deep (15+)—these are pacing estimates, not stopping rules.
3. Is there a previous probe-ledger to calibrate?

If a ledger is present, enter calibration mode instead of restarting the interview.

Run the engine

Read ../probing/SKILL.md and follow its four phases:

1. Unpack
2. Elicit
3. Challenge, only after explicit consent
4. Settle

At the beginning of each phase, state the current phase and what it will accomplish in one sentence.

Before phase 3, ask permission to challenge the user's premises. If permission is declined, skip phase 3 and mark it as not run in the ledger.

End cleanly

Return the probe-ledger, recommend exactly one test for the coming week, and stop. Do not silently continue into a plan, outline, price sheet, or execution roadmap.

Suggested handoff:

> The probe is complete. The plan is intentionally not written yet. Use the locked facts as inputs to the next planning skill, keep the unverified items visibly labeled as assumptions, or run this week's single test and return with the ledger for calibration.

Commands

• /probe-me — start a standard session
• /probe-me light or /probe-me deep — set depth
• /probe-me skip-adversarial — omit the challenge phase
• /probe-me calibrate plus a ledger — update beliefs from observations
• settle or enough — move immediately to the ledger

Non-negotiable boundaries

• Ask one substantive question at a time after setup.
• Give a recommended answer or hypothesis with each question so the user can react, correct, or reject it.
• Ask only for information unique to the user; research public facts instead of asking the user to supply them.
• Use research as pressure-testing input, not as a lecture.
• Never treat an untested assumption as a fact in downstream work.
• Keep the skill stateless; the ledger carries state between sessions.