Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/cubxxw/agent-kit/deepen-designnpx skills add cubxxw/agent-kit --skill deepen-designgit clone --depth 1 https://github.com/cubxxw/agent-kitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/cubxxw/agent-kit/deepen-design)<a href="https://agentmods.dev/skills/cubxxw/agent-kit/deepen-design"><img src="https://agentmods.dev/badge/skills/cubxxw/agent-kit/deepen-design.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00115 | $0.01230 |
| Opus 5 | $0.00057 | $0.00615 |
| Sonnet 5 | $0.00023 | $0.00246 |
| Haiku 4.5 | $0.00012 | $0.00123 |
Grade A, and why
deepen-design scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 178 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Deepen Design
Create a more ownable design, not merely a more decorated one.
Read references/recursive-tree.md before work. If the user is reacting to an unsatisfying prior result, also read references/failure-patterns.md.
Core distinction
Granularity has levels:
- product truth and desired belief;
- brand theorem and narrative metaphor;
- page architecture and tempo;
- section composition and interaction;
- component language;
- typography, color, spacing, borders, shadows, and motion polish.
Always descend in this order. A larger token set, more shadows, or more micro-interactions is not deeper design when levels 1-4 remain unchanged.
Start from rendered evidence
For an existing interface:
- Preserve the current implementation and user changes.
- Capture the current desktop and mobile first viewport.
- Inspect at least two later sections and one meaningful interaction.
- State what is objectively working.
- State why the page is forgettable in five seconds.
- Separate functional gates from taste judgments.
Treat “still mediocre” as a failed design test. Do not defend an earlier score or continue the same direction by default.
Build a bounded binary tree
At each direction-changing node, create exactly two materially different branches. A valid branch changes the visual argument, information structure, or signature interaction. A palette swap is not a branch.
Use beam search:
- expand two branches;
- make both visible at the cheapest useful fidelity;
- compare them pairwise;
- keep the strongest branch and one credible challenger;
- descend one level;
- backtrack when both children inherit the same weakness.
Do not construct every possible leaf. Spend depth on consequential decisions, not combinatorial volume.
Required depth gates
Depth 0: product invariant
Write:
- the visitor;
- the moment of tension;
- what they should understand, feel, and do;
- the product truth that competitors cannot honestly claim in the same way.
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 178 lines · 115 tokens per session scan A 5da4a99ab466
deepen-design is a skill published in the GitHub repository cubxxw/agent-kit (2 stars, last pushed 29d ago), licensed MIT. It adds 115 tokens to every session and 1,230 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
hallmark
Anti-AI-slop design skill for greenfield pages, audits, redesigns, and design extraction from URLs or screenshots. Use when the user asks to build a new app or landing page, wants to redesign something, invokes Hallmark by name, or uses audit/redesign/study.
accessibility
Consolidated accessibility skill entrypoint for WCAG 2.2, ARIA Authoring Practices, cognitive accessibility, Section 508, EN 301 549, design intent verification, and the Accessibility Planner workflow.
desktop-principles
Desktop-specific UX principles - hover states, pointer precision, keyboard shortcuts, multi-window, focus management. Covers macOS, Windows, Linux, web desktop.
taiyi-ui-design
TaiyiForge 第 4 阶段 — UI/UX 契约,产出 UI-DESIGN.md。四端通用。.
accessibility-analysis
服務可達性分析工具箱(accessibility / service coverage)。當用戶說「30km 路網可達」「最近 X 站」「服務範圍」「沙漠」「孤島」「等時圈」「isochrone」「可達性分析」「路網覆蓋」或要新增任何「離 POI 多遠」類型的分析時觸發。整合 mini-taiwan-pulse + taipei-gis-analytics 兩端 SOP,覆蓋三種視覺模式(路網染色 / Polygon 沿路網 / Hex 格點)、三套既有 reference pipeline 對照、模式選擇決策樹、常見坑(Overpass mirror 不穩 / pyrosm 爆 RAM / multi-bucket /…
ui-design
UI 样式修改协作流程(已有界面的视觉层微调)。触发硬条件:页面已经在代码里跑着,改的只是它的视觉表现——布局、间距、颜色、字号、圆角、组件搭配。通过"读代码 + ASCII 画出现状让用户确认 → 给 2-3 个 ASCII 方案 → 用户选定 → 最小改动 → 微调"的流程,减少沟通偏差、避免浪费 token。产出:只动样式的代码 diff。硬边界:不动业务逻辑、不动数据流、不改交互行为、不顺手重构。不适用于:界面还不存在、要从零探索长什么样(用 design-exploration)、改的是功能或交互逻辑而不只是视觉(用 req-change-workflow)、照着设计图/截图复刻整页(用…