product-polish-workflow
Skill0 starsChanged 21 days agoMITView on GitHub
Use when an existing product needs a code or design audit, user-journey QA, UI/UX polish, branding work or a detailed next-feature plan. It keeps scope, evidence, code ownership, the journey tracker and verification, and calls color-kit for palettes and modes, ui-ux-kit for screens, components and motion, design-system-kit for the token system and prototype-wireframe for throwaway answers. Not for building a new palette, a new design system or a single new screen from scratch.
In this repository57 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .cursor/rules/comments.mdcCursor rule
- .cursor/rules/commit-messages.mdcCursor rule
- .cursor/rules/git-workflow.mdcCursor rule
- .cursor/rules/manage-stack.mdcCursor rule
- .cursor/rules/no-worktrees.mdcCursor rule
- manage-stackSkill
- convert-as-skillSkill
- create-agentSkill
- create-agents-mdSkill
- create-commandSkill
- create-hookSkill
- create-ruleSkill
- create-skillSkill
- retrospectiveSkill
- super-architectureSkill
- super-cachingSkill
- super-elixirSkill
- super-nodejsSkill
- super-protobufSkill
- super-tech-blueprintSkill
- atomic-semantic-commitSkill
- code-reviewSkill
- create-bruno-collectionSkill
- create-documentationSkill
- create-ticketsSkill
- ready-specSkill
- write-adrSkill
- write-changelogSkill
- write-mr-descriptionSkill
- write-planSkill
- write-postmortemSkill
- write-runbookSkill
- noslopSkill
- output-saversSkill
- safe-refactorSkill
- tddSkill
- verifySkill
- get-movingSkill
- later-listSkill
- standupSkill
- challenge-meSkill
- codebase-learnerSkill
- researchSkill
- second-opinionSkill
- traceSkill
- color-kitSkill
- design-system-kitSkill
- diagramSkill
- infographicSkill
- prototype-wireframeSkill
- ui-ux-kitSkill
- evidenceSkill
- find-skillsSkill
- handoffSkill
- rumblingSkill
- use-levistackSkill
The file
Files
Product polish workflow
Improve the actual product using current source, observed journeys and the user's scope. The procedures are embedded in this package. Read only the local guides relevant to the task, then work step by step. You may invoke other skills when useful.
| Task | Local procedure |
| Code, architecture, asynchronous state or maintainability | Source audit and boundaries |
| UI/UX polish, accessibility, and who owns palette, system, components and motion | Design and interaction |
| Evidence-based design critique or refine/redesign decision | Design audit |
| Broad feature inventory or end-to-end user-story testing | Journey QA |
Audit motion that already ships (craft is in ui-ux-kit) | Motion |
| Logo, SVG assets or animated identity | Branding |
| Runtime checks and evidence limits | Verification |
| A throwaway prototype to answer a design or behavior question | Prototype, then the prototype-wireframe skill |
| Detailed feature plan or handoff | Planning and delivery |
| Full product-polish result | Minimum result and handoff template |
| Which skill owns which topic | Ownership |
| Optional project examples | Eave, Tabshelf |
Workflow
- Establish scope. Locate the repository, read governing instructions, current Git state, manifests and relevant design/architecture documents, then inspect affected source and callers. Name the audience, primary task, requested mode and acceptance conditions. Audit/plan-only work remains read-only. Preserve unrelated work and current authorization; old branch reports do not authorize Git or release actions.
- Audit the path. Trace entry, state owner, rules and effects. Walk the real journey and required empty/pending/failure/recovery states. Record findings with severity, impact, source, evidence and closure condition. Distinguish reproduced defects, inferred risks and optional improvements. Use one canonical tracker when the request covers many user-visible features. Fix task/data/privacy/state failures before cosmetic drift.
- Choose a bounded direction. Preserve the incumbent platform and useful conventions. Reuse tokens/components. For a novel interaction or architecture choice without precedent, compare two or three small prototypes against the same task before choosing. Inspect an available target before meaningful visual edits. For palette or mode work load
color-kit. For screens, component states, responsive behavior and motion loadui-ux-kit. For the token system loaddesign-system-kit. Use their steps and files, and keep generated concepts distinct from runtime evidence. - Implement the authorized slice. Reproduce behavioral defects first. Simplify at the correct owner; avoid speculative layers, dependencies and migrations. Model record, visit, editor and request lifetimes where needed. When delegation is authorized, give disjoint ownership, evidence, dependencies and acceptance; keep shared assets under one owner. Review specification compliance and quality, then fix and re-review material findings.
- Verify the real result. Derive commands from the project's toolchain. Check behavior, production packaging, rendered states and motion on their matching surfaces. Inspect Light/Dark/System, supported sizes, focus and relevant accessibility preferences. Use isolated fixtures. Record command/observation, source state, outcome and artifact; distinguish passed, failed, skipped and unavailable checks. Historical results are not current proof.
- Leave a concrete continuation. Update the design system and affected docs. Use the minimum result for a full polish engagement and scale it to narrower scope. Keep implemented work separate from proposed features; give requested plans files/contracts, states, dependencies, verification and delivery gates. Check the final diff. Report user-visible results, actual paths, checks, remaining limits and Git state.
Done when
The authorized result is reviewable and its evidence supports the claims. Optional project examples do not constrain other repositories. Profile copywriting, greenfield stack selection and unrelated tasks fall outside this workflow.
Related skills
color-kit: palette, roles, Light and Dark and other modes, contrast proof and stack binding.ui-ux-kit: one screen or component with states, motion, UX and platform behavior. Its acceptance rows join the journey tracker.design-system-kit: the token system andDESIGN.md, then a critique against it. Also extracts a site's design.prototype-wireframe: one throwaway answer to an open question.codebase-learner: map the codebase before a broad source audit.ready-specorwrite-plan: the next feature needs a full plan.tdd: fix a reproduced behavior defect test-first.verify: prove the polished result on the real surface.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

