trellis-brainstorm

Guides collaborative requirements discovery and PRD creation before implementation.

Updated Apr 14, 2026
One-click install
npx skills add https://github.com/exhyy/face-crop --skill trellis-brainstorm-exhyy
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: trellis-brainstorm
Source: https://github.com/exhyy/face-crop/tree/main/.claude/skills/trellis-brainstorm
Command: npx skills add https://github.com/exhyy/face-crop --skill trellis-brainstorm-exhyy

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Starting implementation with vague or incomplete requirements leads to rework, scope creep, and misaligned solutions. This Skill structures the requirements discovery phase so ideas are captured immediately, questions stay high-value, and technical decisions are grounded in research rather than guesswork. ## Core Features & Use Cases - Task-first capture: Creates a task directory and seeds a PRD immediately so no idea is lost during discussion. - Question gating: Filters out derivable or meta questions, asking only one blocking or preference question at a time with concrete multiple-choice options. - Research-first technical choices: Delegates investigation of libraries, patterns, and conventions to a research sub-agent that persists findings to research files. - Diverge-then-converge workflow: Expands thinking across future evolution, related scenarios, and edge cases, then locks an explicit MVP scope with out-of-scope items. - Use Case: When a user describes a new feature like "add a plugin system", the Skill creates the task, inspects the repo, researches comparable plugin architectures, presents 2-3 approaches with trade-offs, and converges on a confirmed PRD with acceptance criteria before any code is written. ## Quick Start Ask the assistant to brainstorm requirements for a new feature you are considering before writing any code.

Frequently Asked Questions about trellis-brainstorm

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

FAQPage Schema
How do I gather requirements before writing code with an AI assistant?▼

Use a structured brainstorm workflow that creates a task directory, seeds a PRD, and asks one high-value question at a time. The assistant inspects the repo first, researches technical options, and converges on an MVP scope with acceptance criteria before implementation starts.

What is a PRD and how do I write one for a coding task?▼

A PRD (product requirements document) records the goal, requirements, acceptance criteria, out-of-scope items, and technical decisions for a task. This workflow seeds the PRD immediately and updates it after every user answer until requirements are confirmed.

When should I skip brainstorming and implement directly?▼

Skip brainstorming for trivial changes like single-line fixes or typos, and for simple tasks with a clear goal affecting one or two files. Brainstorming adds value when requirements are vague, multiple approaches exist, or architectural trade-offs matter.

How does research-first decision making work for technical choices?▼

Instead of asking the user to invent options, the workflow spawns research sub-agents that investigate comparable tools and conventions, persist findings to research files, and return summaries. The main agent then presents 2-3 concrete approaches with trade-offs for the user to choose.

What are the limitations of a brainstorm-only workflow?▼

Brainstorming produces a confirmed PRD and implementation plan but does not write code itself. It also depends on user answers for blocking and preference questions, so progress stalls if the user does not respond to the current question.