short-talk

Structures assistant dialogue into short messages with explicit status and handoff points.

Updated Nov 21, 2025
One-click install
npx skills add https://github.com/alatyshau/duet --skill short-talk-alatyshau
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: short-talk
Source: https://github.com/alatyshau/duet/tree/main/packages/skills/modes/short-talk
Command: npx skills add https://github.com/alatyshau/duet --skill short-talk-alatyshau

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long assistant messages strip the user of control: they go unread, get cited selectively, and lose their status (question, report, or proposal). This Skill keeps the user able to redirect, reject, or drill in at any moment by enforcing short, status-explicit messages. ## Core Features & Use Cases - Word limit enforcement: Default ~100 words per message, configurable by the user at invocation; exceeding the limit requires explicit permission. - Chunking at handoff points: Long content is split only where the user can stop, redirect, or drill in, never mid-thought. - Explicit message status: Every message signals what is expected of the reader — question, report, proposal, observation, or permission request. - Use Case: During a multi-step planning discussion, the assistant delivers one decision point per message and waits for confirmation before continuing, so the user can steer the conversation at every step. ## Quick Start Invoke the short-talk skill and optionally set a word limit, for example: use short-talk with up to 50 words per message for this discussion.

Frequently Asked Questions about short-talk

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make an AI assistant respond in short messages?

Invoke the short-talk skill, which caps each assistant message at roughly 100 words by default. You can set a custom limit at invocation, such as up to 50 words or up to three paragraphs.

How to chunk long AI responses without losing context?

Split content at handoff points where the user can stop, redirect, or drill in, never mid-thought. After each fragment, ask a short continuation question and wait for explicit confirmation before proceeding.

Can the assistant exceed the word limit for long quotes or artifacts?

Yes, but only with permission. If content genuinely cannot be chunked, such as a finished artifact or full schema, the assistant must ask first, for example: dump the whole block, about N words?

What is explicit message status in AI dialogue?

Every message must signal what is expected of the reader: a question to answer, a report to accept or correct, a proposal to approve or reject, an observation to contest, or a permission request to grant or deny.

When should short-talk not be used?

Short-talk does not apply to final artifacts themselves, such as declarations, manifestos, READMEs, or specs. It governs only the discussion around those artifacts, not the deliverable documents.