Product Design Skill: Make Your Agent Decide Interaction Flows

Where this fits

You're building an agent that helps users design a checkout flow, and the user asks: "Should deleting an item be undoable?" Without a clear decision layer, your agent might guess or jump straight to UI styling. The product-design skill fills that gap—it tells your agent how to reason about interaction choices, action scope, reversibility, recovery, and all reachable states before any visual work begins. It's the 'what should happen' step before 'how it looks'.

Why agents benefit

• Decides interaction and control types — the agent can pick between toggles, checkboxes, modals, or inline actions based on context, not vibes.
• Clarifies scope and consequence — every action gets a defined object, scope (e.g., single item vs. bulk), and visible outcome, so the agent avoids ambiguous instructions.
• Settles reversibility and safeguards — it determines whether an action needs undo, confirmation, or a grace period, and which safeguard fits.
• Enumerates reachable states — the agent maps every possible state (loading, empty, error, partial success) to ensure the flow handles reality beyond the happy path.
• Routes to specialized skills — it knows when to hand off to ui-design, ui-animation, or copywriting, keeping the decision layer separate from implementation.

Practical scenarios

Designing a delete action with undo

Your agent is asked to design a file deletion flow. Instead of defaulting to a modal, it uses product-design to decide that deletion is reversible, picks an undo toast with a 5-second window, and defines the states: visible, undone, confirmed, and permanent deletion.

Choosing between a toggle and a checkbox

In a settings screen, the agent must decide how to represent "email notifications." product-design weighs the action's scope (single setting vs. multiple sub-options), reversibility (instant apply vs. save button), and reachable states (enabled, disabled, pending save). It then routes the visual styling to ui-design.

Reviewing a destructive action's confirmation

A user asks: "Should this confirm dialog have a 'Don't ask again' checkbox?" The skill helps the agent evaluate the risk of irreversible data loss, the user's recovery path, and whether the safeguard matches the action's consequence. It outputs a reasoned recommendation, not a guess.

Add it to your agent workflow

Install the skill directly from the agent-skills repository:

npx skills add mblode/agent-skills --skill product-design

Then invoke it in a prompt like: "Use product-design to specify the flow for deleting a project, including reversibility and reachable states." The agent will return a structured decision and route implementation tasks to the appropriate skills.

Compare before adopting

Before adding it, compare against similar decision-layer skills. Check quality signals: does it produce concrete, testable decisions? Look at maintenance freshness—last commit and issue responses. Alternatives like product-judgment (for investment decisions) complement this skill rather than replace it. Verify that it fits your workflow: if your agent already has a strong product reasoning step, this may be redundant.

Why it is worth tracking

With 104 stars and an active author, product-design is a focused, well-scoped skill that fills a real gap in agent-based product design. It's not a flashy UI generator—it's a decision engine that makes your agent's flows more robust and human-centered. Track it if you're building agents that need to reason about interaction design, and evaluate it when your agent starts making product choices beyond simple command execution. For a deeper look, see the skill overview on OpenAgentSkill.