Design

SkillMedia

Creates design artifacts through collaborative brainstorming of approaches, architecture, and trade-offs. Use when the user says "design this", "create a design", "brainstorm approaches", or "write a design doc".

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Design skill

What this skill tells your AI

The instructions your AI receives, as published by codagent-ai/agent-skills in skills/design/SKILL.md and read by ahel’s review.

Turn approved specifications into an implementation-ready technical design. Requirements are already settled; focus on architecture, component boundaries, interfaces, data flow, failure handling, migration, testing, and meaningful trade-offs.

Do not implement, scaffold, or invoke an implementation skill before the user approves the design.

Process

  1. Read every specification, including deferred-to-design markers, and inspect the relevant code, interfaces, tests, and repository conventions.
  2. Use codagent:ask-questions for consequential architectural choices that repository context cannot safely resolve. Evaluate options first, recommend a path, and decide low-risk implementation details yourself.
  3. Compare plausible approaches when a real trade-off exists. Do not manufacture alternatives for an obvious, patterned solution.
  4. Present a design scaled to the change's complexity. Cover the important components, interactions, decisions, risks, verification strategy, and any resulting specification implications; use diagrams when they clarify the design.
  5. After approval, write design.md and apply any specification changes revealed by the design, including completing deferred scenarios, only when those implications were presented with the approved design. Return to the user for a newly discovered behavioral or scope decision. Keep normative behavioral changes in specs and technical rationale in the design.

If design work exposes a product or scope decision rather than a technical implication, discuss it with the user instead of silently inventing behavior. The written artifacts must be self-contained for an implementing agent with no conversation history.

Do not invoke another lifecycle skill after writing.

Artifact template

Omit sections that do not apply.

## Context

<!-- Relevant current state and constraints. -->

## Goals / Non-Goals

**Goals:**
<!-- Outcomes this design enables. -->

**Non-Goals:**
<!-- Explicit exclusions. -->

## Approach

<!-- Components, interfaces, interactions, data flow, and failure behavior. -->

## Decisions

<!-- Consequential choices and rationale. -->

## Risks / Trade-offs

<!-- Material risks, mitigations, and alternatives considered. -->

## Migration Plan

<!-- Rollout and rollback when applicable. -->

## Open Questions

<!-- Only unresolved decisions that remain. -->

Signals

GitHub stars
30
Forks
1
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
design-codagent-ai
Source
github.com/codagent-ai/agent-skills