briefing

Walk through unresolved issues one by one with deep analysis and file-fixated decisions.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Unresolved issues, open questions, and pending decisions often get shallow treatment when batched together, and decisions made in chat are lost when the conversation context is compacted. This Skill enforces a disciplined one-issue-at-a-time walkthrough where each issue receives full analytical depth and every approved decision is materialized into a file. ## Core Features & Use Cases - Index-first workflow: Scans a source (file, conversation, agent reports) to build a lightweight index of unresolved issues before any deep analysis begins. - Structured presentation: Each issue is presented with mandatory Problem, Solution, and Why blocks (or a Fork block for genuine decision forks), validated by a Cold Reader Test covering five questions. - Fixation strategies: Decisions are persisted via one of four strategies (log, artifact, final, chat) recorded in a briefing file header, so knowledge survives session compaction. - Use Case: After a design review produces 5 open questions about a system prompt, run a briefing to walk through each question sequentially, get explicit user approval per decision, and record all reasoning in briefing_<case>.md linked from plan.md. ## Quick Start Ask the agent to run a briefing on the open issues in your review document and walk through them one at a time.

Frequently Asked Questions about briefing

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

FAQPage Schema
How do I review multiple open issues without losing depth on each one?

Process issues one at a time instead of batching them. Build an index of titles and pointers first, then give each issue a full analysis with Problem, Solution, and Why blocks before moving to the next only after explicit approval.

How to preserve decisions made in AI chat sessions across compaction?

Fixate every approved decision into a file using a chosen strategy: log decisions in a briefing file, write them into a design artifact, or apply them directly to code and specs. A status label in an index alone is not fixation.

What is the Cold Reader Test for decision briefings?

It is a self-check where you reread your briefing as someone with zero context and verify five questions are answerable from the text alone: the problem, the proposed solution, why it solves the problem, why it beats alternatives, and what makes it excellent.

When should I present a fork instead of a single recommendation?

Use a fork only when you have done all the analysis possible and the remaining choice genuinely depends on a user priority you cannot know. Present the options with consequences and state which priority settles the choice.

Which fixation strategy should I choose for interconnected issues?

Choose the log strategy when issues are interconnected: record all decisions with reasoning in the briefing file first, then update the target document in one coherent pass at the end to avoid rework from changing context.