Build throwaway prototypes to answer design questions about logic or UI.
npx skills add https://github.com/sanjanb/my-agent-harness --skill prototype-sanjanb
Please help me install this Agent Skill. Skill: prototype Source: https://github.com/sanjanb/my-agent-harness/tree/main/skills/mp-prototype Command: npx skills add https://github.com/sanjanb/my-agent-harness --skill prototype-sanjanb
?variant= URL param and a floating bottom bar with arrow-key navigation. - Capture Workflow: Fold validated decisions into real code while preserving the prototype on a throwaway branch as a primary source. - Use Case: You are unsure whether a state machine handles a tricky edge case. The Skill builds a one-command terminal app where you press keys to push the model through real transitions and immediately spot the flaw. ## Quick Start Ask the agent to prototype whether your state model handles a specific edge case, or to generate three different layout variants for an existing page.High-intent search queries and answers about installing and using this skill.
Build a small terminal app that wraps a pure reducer or state machine, rendering the full state after every keystroke. Keep the logic module free of I/O so it can be lifted directly into the real codebase once validated.
Render each variant on the same route gated by a ?variant= URL search param, with a floating bottom bar to cycle between them. Keep existing data fetching intact so variants are judged against real content and layout.
No. A prototype that needs tests is no longer a prototype. Skip tests, persistence, and polish; the goal is to answer one design question fast, then rewrite the winning approach properly when folding it into production.
Use a logic prototype when the question is whether a state model or data shape behaves correctly under edge cases. Use a UI prototype when the question is what a page or component should look like, comparing structurally different layouts.
Fold the validated decision into the real code, then commit the full prototype to a throwaway branch outside main as a primary source. Leave a pointer to that branch on the implementation issue so the context survives.
letta-aiSubmits user-approved agent feedback reports to the Letta team via CLI.
nklischGenerate production-ready code from design specifications with tests.
wnz99Refactor Python code into clearer, smaller units with preserved behavior.
zircoteCreate executive ADRs with sections for SWOT analysis, costs, benefits, risks, and sign-off.