Wait What

---
name: wait-what
description: 'Use when the user says "wait, what", "the explanation is unclear", "say that again", or asks to restate the last response in plain language. Re-pitches the previous explanation with the missing frame supplied, in plain English and the project''s ubiquitous language. Don''t use for tasks that require source or remote-system changes.'
---

# Wait, what

## Contract

| Field | Bound contract |
|---|---|
| Trigger | User says 'wait, what', 'I don''t follow', or 'say that again'. |
| Authority | Read-only: no file, VCS, credential, paid, published, deployed, or remote mutation. |
| Side effect | Restates the same claim without softening; nothing touches disk. |
| Done | The claim is re-pitched with the missing frame supplied, in plain English and the project''s ubiquitous language. |

## Inputs

The trigger identifies which explanation failed. No external files are required. Read `CONTEXT.md` at the repository root (or the per-context `CONTEXT.md` beside the relevant source when the project keeps a `CONTEXT-MAP.md`) to use the project''s ubiquitous language. If no glossary exists, use the names already in the code.

## Entry phrasings

One skill, three entry phrasings, one mechanism — re-pitch the same claim with the missing frame supplied in plain language:

- "wait, what" or "say that again": the explanation did not land. Re-pitch the same claim, preserving its substance and conclusions, and lead with the context the explanation assumed.
- "the explanation is unclear" or "context was omitted": the prior explanation skipped background, definitions, assumptions, causal links, or prerequisite steps. Re-pitch from scratch in plain language, embedding the missing context at the point where the original jumped ahead.
- "restate in plain language" or "bro": restate the last response concisely without jargon. Preserve the original meaning and add no new claims, opinions, or information beyond the source response.

All three produce one plain-language re-pitch in chat. None touches disk.

## Procedure

1. Identify the explanation that did not land: the last assistant message immediately before the trigger. Done when: the failed explanation is identified.
2. Restate the same claim, preserving its substance and conclusions. Do not soften, hedge, or replace it with an easier version. Done when: the claim is restated with substance and conclusions intact.
3. Re-pitch it: lead with the context the explanation assumed (one or two sentences on where the conversation is and what the message was answering — the gap is a missing frame, not a missing word); write in ISO 24495-1 English (short sentences, active voice, direct address, common words replacing jargon); use the project''s ubiquitous language. Done when: the claim is re-pitched with the missing frame supplied in plain English and ubiquitous language.

## Failure and recovery
Non-converged: the user signals they do not understand after one re-pitch. Stop. Do not elaborate, restate a third time, or widen scope. The user is the authority on whether the explanation landed. Report `non-converged`.
- No prior explanation or response: if no prior assistant message exists to re-pitch or restate, stop and report that there is nothing to restate. Do not invent a prior explanation.
- Ambiguous confusion signal: if the user's indication is unclear about which part was hard to follow, ask which part of the prior explanation was unclear before re-pitching.
- Scope creep: if the re-pitch would require teaching an entire prerequisite topic, state the prerequisite boundary and offer to cover it separately. Do not widen into a tutorial.
- Clean re-pitch: keep the re-pitch free of caveats, meta-commentary, or references to the original explanation's failure.

## Output
A single re-pitched message in the conversation, in plain English with the missing context supplied.