steering-interview

Captures author input before research and after briefs to steer content generation.

1|Updated Apr 9, 2026
One-click install
npx skills add https://github.com/AlexYedi/Empire_State_Events_Pipeline_Take_3 --skill steering-interview-alexyedi
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: steering-interview
Source: https://github.com/AlexYedi/Empire_State_Events_Pipeline_Take_3/tree/main/.claude/skills/steering-interview
Command: npx skills add https://github.com/AlexYedi/Empire_State_Events_Pipeline_Take_3 --skill steering-interview-alexyedi

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Content pipelines often generate drafts without the author's freshest perspective, forcing corrections afterward. This Skill inserts two structured interview touches into any content flow so the author's intent steers research direction and editorial decisions before drafting begins. ## Core Features & Use Cases - Touch 1 (Aim): A light, skippable five-question intake run before research that routes answers on content, format, additional research, context, and audience into the research fan-out and generation prompts. - Touch 2 (Sharpen): A collaborative gate after the research brief that presents up to three genuine editorial forks derived from the brief, letting the author decide before any draft exists. - Persistence & Routing: Captures answers verbatim into an Author Steer block appended to the event or content page, threading them into downstream prompts and later mining them to update voice and style guides. - Use Case: Before drafting a post-event recap, run Touch 1 to learn the author wants deeper research on a specific speaker, then after the brief surfaces a tension between two speakers' positions, run Touch 2 to decide whether to lead with the conflict or pick a side. ## Quick Start Say "steer this" or "interview me first" before starting any content deliverable to run the two-touch steering intake.

Frequently Asked Questions about steering-interview

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

FAQPage Schema
How do I add human input to an AI content generation pipeline?

Use a two-touch interview pattern: a light pre-research intake that routes answers into research scope and content prompts, plus a post-brief gate presenting genuine editorial forks before drafting. This keeps author input ahead of generation instead of as corrections afterward.

How do I steer AI research before it starts?

Ask a single batched set of open-ended questions covering content, format, additional research targets, context, and audience before launching research agents. Route the research-specific answer directly into the fan-out prompts so it shapes what gets investigated.

Can this interview run as a subagent or background task?

No. The interview is multi-turn and human-in-the-loop by definition, and subagents are one-shot and non-interactive, so it must run in the main conversation thread. Only the protocol—questions, capture schema, routing rules—is reusable.

What happens if the research brief has no real decisions to make?

The Sharpen touch is skipped cleanly with a one-line note that the brief is unambiguous, and drafting proceeds immediately. Manufactured questions are explicitly treated as worse than asking none.

How are interview answers stored and reused?

Answers are captured verbatim into a structured Author Steer block appended to the event or content page, threaded into downstream prompts in the same run, and later mined to update voice and style guide files.