How does Clarify determine if a decision needs human input?

Decisions are classified as Tier 3 (Human required) if the codebase evidence is silent, web research is conflicting, or the choice is inherently subjective.

Does this skill help with implementation or just planning?

Clarify is specifically designed for the pre-implementation phase to ensure that the plan is explicit and agreed upon, which prevents 'churn' and rework during the actual coding phase.

Can Clarify leverage my existing project context?

Yes, it automatically scans for existing session files and workspace context to assess how much clarification is needed, skipping unnecessary research steps if the goal is already clear.

What is the difference between Default and --light mode?

Default mode performs deep, autonomous research using sub-agents and domain experts to resolve ambiguity, while --light mode focuses on quick, iterative Q&A with the user without launching extra research agents.

Clarify Requirements

Name: Clarify Requirements
Author: corca-ai

bycorca-ai

•

Productivity & Workflow

Resolves technical ambiguity and refines project requirements through deep codebase research and multi-perspective expert analysis.

The Clarify skill for Claude Code minimizes implementation churn by transforming vague requests into explicit, actionable plans. By leveraging autonomous sub-agents for codebase and web research alongside domain-expert analysis, it identifies critical decision points and classifies them by evidence source. Whether using the deep research-first default mode or the iterative --light mode, Clarify ensures that developers and AI agents start every task with a clear scope, defined constraints, and resolved technical conflicts before a single line of code is written.

Key Features

01Autonomous codebase and web research sub-agents for deep context gathering.

0273 GitHub stars

03Session-aware context recovery for continuous requirement refinement.

04Persistent 'Why-dig' questioning to uncover hidden technical ambiguities.

05Multi-perspective domain expert analysis to evaluate architectural decisions.

06Tiered decision classification to separate autonomous fixes from human-required choices.

Use Cases

01Resolving conflicting architectural patterns in complex legacy codebases.

02Establishing BDD-style success criteria before starting major implementation tasks.

03Refining vague feature requests into detailed technical specifications.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add corca-ai/claude-plugins clarify

For use in Claude.ai and ChatGPT

Download Skill