| name | unpack |
| description | Unpack complex work or a dense topic one idea at a time, checking comprehension at each step rather than delivering a wall of text. Use when the operator asks to "unpack", be "caught up", or "walked through" something — and offer it proactively before any recap, plan, or explanation long enough to overwhelm in a single message. |
| argument-hint | [optional: the topic or work to unpack] |
Unpack — one idea at a time, until it lands
Transfer understanding the operator could explain back to someone else — not the feeling of
being informed. You cannot transfer understanding by volume: the wall of text delivers
awareness and no comprehension. You transfer it one rung at a time, and the operator does the
cognitive work — you feed the right idea at the right altitude, then stop and confirm it
landed before adding the next. Never trade their comprehension for your speed.
Step 1 — Build the ladder (privately, before you say anything)
Decompose the topic into a dependency-ordered sequence of single ideas, each resting on the ones
before. Order by understanding, not chronology or how you built it. Keep it to ~5–8 rungs; more
means you're mixing altitudes — collapse detail upward. A default skeleton (adapt freely):
- Why it exists — the problem, as a plain mental model. (Almost always start here.)
- The core tension — the one hard thing that makes this non-trivial.
- The mechanism — how it works, high level.
- The constraints — the guardrails that must hold.
- The evidence — what was done / proof it works.
- The scrutiny — how it was stress-tested, or where it's genuinely uncertain.
- The honest gaps — what's not done. Glossing edges produces false confidence, worse than a
known unknown ([[no-fabrication]]).
Step 2 — Frame it, then find where they already are
Open with how this goes: layer by layer, one idea at a time, you'll stop and check, they set the
pace. Then locate their starting rung — ask ("how are you thinking about this now?") or deliver
rung 1 and invite "skip ahead if you know this." Never assume their start; never start above it.
Step 3 — The per-turn loop (the heart of the skill)
For each rung, one message:
- One idea. Only one. The "and also…" you want to add is the next rung. Hold it.
- Plain language. Translate every term; name the technical word in parentheses after the
plain version, never before.
- Anchor to what they know. Reach for an analogy from their world — their business, a process
they run, something established earlier — not a generic one.
- Then STOP on a check that makes them work — a question ("in your own words, why X?"), a
reaction gate ("does this land, or is anything fuzzy?"), or a predict ("what would you expect
if…?"). The check is what makes them follow; skipping it because they seem to be keeping up
defeats the skill.
- Wait. The next rung does not appear in this message.
Step 4 — Calibrate to their response, every turn
- "got it" / "next" / "skip" → seal the prior point in one line so it isn't lost to speed, then
advance one rung. Don't re-teach.
- A wrong or partial check answer → re-explain that rung a different way (new analogy, smaller
piece), never the same explanation louder. Stay on the rung until it lands.
- A question → answer at their altitude, then resume where you were.
- "too much" / overwhelmed → you went too wide or deep. Drop to a smaller idea.
Step 5 — Land it
After the last rung, give a compact recap of the full ladder in one tight list — a consolidation of
rungs already understood, not a fresh dump — so the layers fuse into one structure they can hold.
Then ask if any layer is still fuzzy before you call it done.
Proactive offer (don't auto-run)
When you're about to dump a wall of text — a multi-step recap, a dense plan, a hairy bug, an
architecture — offer first: "This is a lot — want me to walk you through it one layer at a time so
it lands?" If they'd rather have it all at once, give it. Offer, don't impose.
