spec-builder

Generate typed per-module spec shape files with YAML frontmatter via structured interviews.

1|Updated Jan 19, 2026
One-click install
npx skills add https://github.com/saxonthune/carta --skill spec-builder-saxonthune
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: spec-builder
Source: https://github.com/saxonthune/carta/tree/main/.claude/skills/spec-builder
Command: npx skills add https://github.com/saxonthune/carta --skill spec-builder-saxonthune

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It helps you elicit product or system requirements from users and turn them into well-partitioned, machine-readable spec shape files, avoiding vague notes and incomplete requirements.

Core Features & Use Cases

  • Structured requirement interviews: guides the user through seed → draft-with-gaps → targeted interrogation to resolve missing details.
  • Typed spec shape files per module: outputs one markdown file per module with YAML frontmatter (e.g., id, type, feature, depends-on, exposes), plus typed sections for inputs/outputs/errors/business rules.
  • Smell- and completeness-aware elicitation: explicitly tracks open questions and flags vague/weak/unresolved areas before final writing.
  • Use cases: spec out an onboarding flow, define a new feature/module/service, or build a spec group directory for a Carta workspace.

Quick Start

Ask the AI to run spec-builder for your account-creation module by saying: Interview me about the requirements for the account creation flow and write the spec shape files into my .carta/ workspace.

Frequently Asked Questions about spec-builder

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

FAQPage Schema
How do I generate typed module spec files from product requirements?

Typed module spec files are generated by running a structured interview that elicits product requirements, fills gaps, and outputs markdown files with resolved YAML frontmatter and typed sections for inputs, outputs, and business rules.

What is the best way to elicit requirements for a new feature without missing edge cases?

Requirements elicitation is best handled through a seed, draft-with-gaps, and targeted interrogation workflow that tracks open questions and explicitly flags vague or unresolved areas before writing the final specification files.

How do I structure a service specification so dependencies and inputs are machine-readable?

Service specifications are structured using YAML frontmatter with fields like id, type, depends-on, and exposes, ensuring dependencies, inputs, outputs, and errors are explicit and parseable for downstream tooling.

Can I write spec shape files directly into a Carta workspace directory?

Yes, spec shape files can be written directly into a user-specified output location such as a Carta workspace directory by instructing the tool to save the generated markdown files there during the interview process.

Does spec generation prevent the AI from inventing business rules that were not discussed?

Yes, spec generation prevents invented business rules by only writing resolved fields based on user input, using TBD placeholders only when necessary, and explicitly tracking open questions during the elicitation process.

When should I avoid using an LLM interviewing approach for documentation workflows?

You should avoid LLM interviewing for documentation workflows when your project requires zero TBD placeholders, when you lack a defined module or service boundary, or when business rules cannot be explicitly stated.