Install
$ agentstack add skill-bnet47-codexicon-design-experience ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README — it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming — see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps — measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Design experience
Create a coherent customer experience, not a decorated wireframe. Make every visual choice serve the product, audience, setting, and task.
1. Establish context
Read the approved charter or feature brief, existing interface, design system, brand assets, content, and relevant conventions. Read agent_docs/design.md when it exists. Do not create durable design context unless the user requests it or the approved workflow requires it.
Define before implementation:
- user, primary task, setting, and observable success;
- surface register: brand when design carries identity and persuasion, or product when design primarily serves repeated work;
- product personality, emotional tone, and anti-references;
- one memorable, context-specific design idea;
- brand, platform, accessibility, performance, content, and licensing constraints;
- required viewports, states, and interaction paths.
Write one physical-scene sentence: who uses this, where, in what conditions, and with what level of urgency or attention. Use it to choose light versus dark, density, contrast, target sizes, motion, and information hierarchy.
Set three dials explicitly: variance (quiet to expressive), motion (still to dynamic), and density (spacious to compact). Do not equate distinctive with loud.
2. Inspect before inventing
- Reuse established tokens, components, content style, and asset language when they work.
- Inspect the running product and representative project files; source code alone does not reveal the experience.
- For an existing surface, preserve product identity unless the request explicitly authorizes redesign.
- Verify current platform guidance and references when they may have changed. Never clone proprietary layouts or assets.
- Confirm fonts, icons, images, and libraries are available, performant, and licensed.
For visual-system work, new surfaces, redesigns, or polish, read [references/interface-craft.md](references/interface-craft.md). For production flows, forms, onboarding, or release hardening, also read [references/hardening.md](references/hardening.md).
3. Shape before styling
Define the information order, primary action, navigation model, state transitions, content requirements, and responsive behavior before choosing decorative treatments. Remove sections and controls that do not advance the task or argument.
State the direction in one sentence, then encode it consistently through typography, color, spacing, shape, imagery, motion, and information density. Use real product language. If real content is unavailable, mark neutral placeholders clearly and avoid invented proof or business facts.
Run both reflex checks:
- Could the category alone predict the palette and layout?
- Could the category plus a request to avoid common clichés predict the replacement aesthetic?
If either answer is yes, return to the scene, content, and product artifacts instead of swapping one fashionable template for another.
4. Implement the complete experience
- Preserve working behavior and build the smallest complete flow.
- Cover every reachable default, hover, focus, active, disabled, loading, empty, error, success, stale, and permission state.
- Make responsive behavior deliberate at content-driven breakpoints.
- Use semantic structure, keyboard operation, visible focus, usable targets, sufficient contrast, reduced-motion support, and helpful validation.
- Keep navigation, state, and failures observable and recoverable. Preserve deep links and browser behavior when applicable.
- Stress long text, localization growth, narrow screens, zoom, slow networks, missing data, and repeated content.
- Avoid decorative complexity that harms loading, readability, maintainability, or task completion.
- Never fabricate product data, customer quotes, logos, outcomes, social proof, or decorative dashboards.
Use an installed specialist site, browser, image, document, or presentation skill when the requested artifact needs it. Follow that skill's render-and-verify workflow.
5. Iterate on rendered evidence
Run the real interface and inspect it at agreed critical widths. Exercise the primary flow plus loading, empty, error, keyboard, zoom, and reduced-motion behavior.
Use two passes:
- Composition: hierarchy, rhythm, typography, color, imagery, density, responsive composition, and unwanted template reflexes.
- Operation: comprehension, interaction feedback, recovery, accessibility, content stress, and performance.
Fix the highest-impact gap, rerender, and recheck. Do not declare visual quality from code inspection. If rendering is unavailable, report the limitation.
6. Report
Summarize the design direction and dials, implemented behavior, rendered and functional verification, and residual constraints. Recommend $review-creative for an independent quality pass on customer-facing or high-impact work.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: bnet47
- Source: bnet47/codexicon
- License: MIT
- Homepage: https://bnet47.github.io/codexicon/repo-template-playbook.html
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.