requirement-convergence

Converges requirements into outcome, scope, non-goals, and cost bands before design begins.

675|103|Updated Oct 10, 2025
One-click install
npx skills add https://github.com/shinpr/claude-code-workflows --skill requirement-convergence
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: requirement-convergence
Source: https://github.com/shinpr/claude-code-workflows/tree/main/dev-workflows/skills/requirement-convergence
Command: npx skills add https://github.com/shinpr/claude-code-workflows --skill requirement-convergence

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Requirements often arrive bloated, ambiguous, or aimed at the wrong outcome, causing AI agents to faithfully build the wrong thing. This Skill separates the outcome a change must produce from the requirements proposed to reach it, records user-authored exclusions, and bands cost from structural evidence before any design work starts.

Core Features & Use Cases

  • Convergence Fields: Defines four fields (outcome, requirements, nonGoals, cost) with explicit pass conditions and readiness labels (ready, weak, weak-but-explicit) so scope is judged, not assumed.
  • Hearing Protocol: A five-step elicitation process that presents scope facts first, asks at most two questions per message, and records answers in the user's own wording to keep decisions reviewable.
  • Cost Banding from Structure: Estimates cost bands (low-reversible, medium, high-irreversible) from grep/glob evidence about targets, boundaries, reuse, and migrations rather than behavior reading, and maps each band to a challenge intensity.
  • Use Case: Before designing an account-recovery flow, run this Skill to pin down the single observable outcome, label each requirement as current-state or desired-future, capture what the user explicitly excluded, and flag any high-irreversible cost drivers needing a user decision.

Quick Start

Ask the agent to run requirement convergence on a new feature request before any design work begins.

Frequently Asked Questions about requirement-convergence

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

FAQPage Schema
How do I clarify ambiguous requirements before starting design?

Run a convergence pass that separates the single observable outcome from the requirements list, labels each item current-state or desired-future, and records user-authored non-goals. Each field gets a readiness label, and design only starts when all applicable fields are ready or explicitly accepted as weak.

How to estimate change cost without reading all the code?

Band cost from structural evidence: count targets and boundaries via grep/glob, check for existing equivalents, detect schema migrations, and verify test harness existence. The result is a band (low-reversible, medium, high-irreversible) with its unknowns, not a person-day estimate.

What is the difference between current-state and desired-future requirements?

Current-state requirements describe behavior that already exists and are facts, not work. Desired-future requirements are the actual change being requested and are buildable. Mixing them into one list makes bloated plans look coherent, so each item must be explicitly labeled.

When should a requirement be challenged before acceptance?

Challenge intensity follows the cost band: low-reversible changes are recorded and accepted, medium-cost changes get one lower-cost alternative presented, and high-irreversible changes (public contracts, persisted data shapes, dependency swaps) require an explicit user decision before design.

What happens to requirements left unresolved during convergence?

A field that still fails its pass condition after one re-ask can be marked weak-but-explicit only with the user's agreement. Downstream consumers treat it as a recorded open question and return the missing decision to the owning workflow when work depends on it.