Skip to content
product-skill
Senior product judgment · for any AI

Turn any AI into a senior product mentor — with one paste.

Generic answers become the kind of advice a seasoned product leader gives: a clear recommendation, the right framework for the moment, and a gut-check on your blind spots. This guide makes the whole skill visual — 30+ frameworks across 10 disciplines, the way a practitioner actually uses them.

See the difference — same AI, same question

YOU ASK“We have 5 features requested for next quarter but only room for 2. How do I decide?”

Plain AI

“It depends on impact and effort. You could try a framework like RICE or MoSCoW. Gather stakeholder input, weigh each feature against your goals, balance quick wins against strategic bets…”

Lists options. Hands the work back to you. No recommendation, no opinion.

AI + product-skill

Use RICE here — you have rough reach and effort, so it gives a defensible ranking fast.

Ship A and C this quarter. A is high-reach, low-effort. C is the enabler the rest depends on. Defer B (high effort, unproven), park D + E.

Blind-spot check: RICE rewards this quarter's reach — if B is your enterprise bet, it'll quietly under-rank it. Flag it as a strategic call, separate from the score.

The A–E features are a placeholder to show the format — not a real case.

Scroll to learn the system

The operating loop

For any product task, it runs the same five moves.

Not a menu dump — a disciplined sequence that ends in a decision you can defend.

  1. 01

    Clarify only if ambiguous

    Ask up to five sharp questions when the request is genuinely under-specified — otherwise proceed. No interrogation.

  2. 02

    Pick the one right framework

    Match the method to the decision — don't dump the whole toolbox on the table.

  3. 03

    Output a decision-ready artifact

    Recommendation first, reasoning second. A scoring table, an eval plan, a PRD — not an essay.

  4. 04

    Tag confidence

    Every substantive call carries a label — from VERY HIGH to SPECULATIVE — so you know how hard to lean on it.

  5. 05

    Run the blind-spot check

    Full ritual for high-stakes, irreversible calls; a one-line tag otherwise. Drop the ritual, never the thought.

Then the next task runs the same five moves — the loop, not a checklist.

The router

What are you trying to do right now?

Tap a task — the skill routes it to the one discipline that owns the answer, then hands you the page.

The task router needs JavaScript to run. Or jump straight in from the Field Guide — every discipline is one hop away.

Start here

Twelve moves that carry most of the judgment.

If you read nothing else, read these. The canon the skill reaches for first — each links straight to where it lives.

01

RICE + kill rule

Reach × Impact × Confidence ÷ Effort. Confidence is an honest discount — kill anything under 50% rather than rank it.

Prioritization →
02

JTBD job story

“When I [situation], I want to [motivation], so I can [outcome].” Frame the struggle, not the feature.

Discovery →
03

Opportunity Solution Tree

Outcome → opportunities → solutions → assumption tests. Every solution ladders back to a real outcome.

Discovery →
04

DHM moat

Delight · Hard-to-copy · Margin-enhancing — the three tests a durable strategic bet has to pass (Biddle).

Strategy →
05

North Star + guardrail

One value-delivery metric + 3–5 movable inputs, paired with a guardrail that must not degrade.

Metrics →
06

AARRR

Acquisition, Activation, Retention, Revenue, Referral — trace a problem upstream to where it really starts.

Metrics →
07

Dunford 5-step positioning

Competitive alternatives → unique attributes → value → target market → market category, derived in order.

GTM →
08

Price before product / WTP

Have the willingness-to-pay conversation during design. No WTP evidence, no spec (Ramanujam).

Pricing →
09

Evals-as-QA loop

For any LLM feature the eval suite is the spec. Analyze → measure → improve, and re-run every change.

AI-Native →
10

Minimum Viable Quality

Set a per-feature bar in three tiers — do-not-ship · acceptable · delight — with a cost envelope up front (Nika).

AI-Native →
11

Type-1 / Type-2 reversibility

One-way doors need pre-mortems and staged approvals; two-way doors just need a fast, reversible call.

Heuristics →
12

Playing-to-Win cascade

Five linked choices: winning aspiration → where to play → how to win → capabilities → management systems.

Strategy →