Raven is an open-source agent harness for running long-term AI work with terminal execution, tracing, memory, skills, evaluation, and reusable workflows. People use the current release to operate and improve persistent AI workflows, while its described future direction is a multi-agent system that combines specialized harnesses.
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 skills add EverMind-AI/Raven --skill visual-artifact-designgit clone --depth 1 https://github.com/EverMind-AI/RavenWrote 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/evermind-ai/raven/visual-artifact-design)<a href="https://agentmods.dev/skills/evermind-ai/raven/visual-artifact-design"><img src="https://agentmods.dev/badge/skills/evermind-ai/raven/visual-artifact-design/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/evermind-ai/raven/visual-artifact-design"><img src="https://agentmods.dev/badge/skills/evermind-ai/raven/visual-artifact-design.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00081 | $0.04580 |
| Opus 5 | $0.00041 | $0.02290 |
| Sonnet 5 | $0.00016 | $0.00916 |
| Haiku 4.5 | $0.00008 | $0.00458 |
Grade A, and why
visual-artifact-design scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 236 lines — stays where its author put it; the contents beside it link to each section on GitHub.
视觉产物设计底座
只在用户最终会看见、阅读、操作或制作某个产物时应用本 Skill。纯后端、数据管道、 Shell 或不产生视觉消费者的文本任务不要建立视觉流程。
本 Skill 只定义共享阶段、证据边界、能力选择和消费者验收。主要领域 Skill 拥有专业
语义、自然母版和领域门槛;$build-polished-visual-frontends 只在 Web 是真实消费者或
合同交付时负责前端技术与最终像素。Runner 只记录传输、工具、日志、用户配置、进程、
客观运行状态和发布事实,不裁决视觉方向、审美质量或设计阶段。文档不能代替运行时事实,
运行状态也不能代替设计判断。
1. 先建立最小合同
选择 Create、Edit 或 Diagnose/Audit。Diagnose/Audit 默认只读,除非用户同时授权
修复。开始制作前,把以下字段写入 Task State 或等价工作记录;只写会改变结果的内容:
primary_domain: 一份主要领域 Skill
natural_medium: 用户实际需要的介质与可编辑母版
consumer: 实际接收、阅读、操作或制作产物的人/软件/设备/流程
target_contexts: 尺寸、设备、状态、观看或生产条件
required_outcomes: 用户必须取得的内容、行为与输出
non_goals: 未承诺、不可验证或不应伪造的范围
acceptance_checks: 每项明确要求对应一个可观察检查
HTML、SVG、图表、地图、控件或动画只是实现形式,不能决定主要领域。若移除某领域的 核心对象、关系或任务价值后产物失去意义,该领域才是主要领域。
既有身份盘点(合同的一部分)
产物有明确主体(产品、项目、组织、活动、人物)时,合同多一行 identity_inventory,在形成任何方向前填好:
到主体自己的官方出处——用户给的材料、代码仓库(README banner、图标目录)、官网 favicon / og 图、代码托管与
社交平台的组织头像(GitHub 固定可取 https://github.com/<org>.png)——逐一查看已经在用的标志、字标、色、字、
图像,记下文件与来源。判定只看像素不看格式:banner 里的字标、位图、截图都是既有标识,"没有 SVG"不等于"没有"。
web_fetch 只回文本,图像用 exec 下载进任务目录再 read_file 查看。
结果只有两种,没有中间地带:
- 找到 → 必须用它:按
references/svg-and-vector.md「Existing marks: three rungs」取用——先找真文件,再从纯色底 原图色键抠出(简单平面标再描 SVG,渐变 / 立体 / 多色标就用抠出的 PNG),只有没有纯色底来源或原图太小时才用 edit 重生成并与原图并排比对;不重画、不另造、不"优化";重做只在用户明说"重新设计 / 换标"时发生,"旧视觉可替换"是模型 自封的分类,不算授权。 - 没找到 → 必须设计出来:标识槽位不留空、不用纯文字或占位顶替、不以"用户没给"跳过;新标识走领域 Skill 的 生成 → 描摹流程,并在 README 写明它是本次新建。
主体的标识与其上级 / 出品方的标识分两层,各归各位:署名位("by X"、页脚、开源角标)用出品方的官方标,主体位
不借出品方符号;主体只有字标没有符号时,favicon / 头像优先直接提取已有字形并注明来源。
直接提取可以裁切、等比缩放、适配画布,不能增造轮廓。若需要新形状,转品牌 Skill,以已有品牌资产作为
image_generate(images=[...]) 的参考生成/编辑候选,再从选定母图矢量化;不由模型手写几何替代。
2. 使用轻量工作状态,而不是完成度叙事
Create 与建立新方向的 Edit 可使用以下状态组织工作。模型可按任务规模合并或跳过确实 不适用的状态,不为走流程制造文书或额外迭代:
CONTRACT
→ REFERENCE_LOCKED
→ VISUAL_THESIS_LOCKED
→ MASTER_ASSET_LOCKED
→ REPRESENTATIVE_FRAME
→ DIRECTION_ACCEPTED | DIRECTION_PROVISIONAL
→ FUNCTIONAL_BUILD
→ SURFACES_CLOSED
→ FINAL_VERIFIED
→ DELIVERED
CONTRACT:第 1 节字段成立,来源、假设与非目标分开。REFERENCE_LOCKED:用户参考、品牌/平台约束或实际观察的生产标杆已转成少量可见 不变量;模型生成的理由仍是候选假设。当方向依赖模板、标杆、品牌或用户参考时,必须记录 实际来源帧、可观察关系、允许迁移轴、不可迁移项、实现目标与终态验证目标;名称和风格标签 不算参考证据。VISUAL_THESIS_LOCKED:主要领域或 companion Skill 要求的视觉命题已写明具体世界、张力、 三秒记忆点,以及它在每个顶层可见区块或页面族中的最终像素落点;形容词、题材联想和只列 少数亮点不算锁定。无需主题性视觉的非 Web 产物或纯工作面在合同中记录为不适用,不伪造状态。MASTER_ASSET_LOCKED:需要主视觉、主场景或其他主导资产时,已选唯一 route、真实取得并查看 母版,锁定语义工作、版面接口、裁切和真实性边界。内容型/营销型 Web 依 companion Skill 要求 在布局前完成主视觉,并为每个主要页面族锁定dominant_background / integrated_visual_field / content_first之一、资产 lineage 与最终证据;占位图不能进入此状态。REPRESENTATIVE_FRAME:用真实内容或明确标注的确定性 fixture,先做一个决定成败的 目标画面或等价终态;不要先铺完整页面、状态和文档。参考合同逐条指向当前画面位置,主要 图像角色使用已经路由并实际取得的资产,不用占位图、手写 SVG 或 CSS 装饰先定版。DIRECTION_ACCEPTED:用户或独立 reviewer 对当前帧和明确范围给出通过。DIRECTION_PROVISIONAL:独立判断不可用但运行模式允许继续;必须标记SELF_REVIEW_ONLY,不能宣称独立视觉通过。FUNCTIONAL_BUILD:方向成立后才扩展合同要求的功能、状态、尺寸与异常路径。SURFACES_CLOSED:每个可见区块、页面族或内容单元都以真实内容完成职责、层级、细节和适用 目标环境检查;所有真实控件与可能被理解为交互的视觉信号均已逐项归类,在目标消费者中兑现并 验证,或改成明确的静态表达。代表帧、构建成功和整幅缩略图不能代替逐区结账。FINAL_VERIFIED:最后一次可见修改之后已重新构建,并在真实消费者中检查当前输出;依赖 参考的任务再次逐条结账可观察关系,未落实项只有硬约束证据才能偏离。DELIVERED:请求的母版、派生物、入口与限制均已交付;发布或 canonical 不从本阶段推导。
What ships with it
10 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.
- examples/chart-page.md 8.6 KB
- examples/computed-geometry.md 11 KB
- examples/design-brief.md 14 KB
- examples/svg-scene.md 9.3 KB
- references/assets-and-imagegen.md 18 KB
- references/data-visualization.md 2.0 KB
- references/evidence-governance.md 7.4 KB
- references/fixed-media.md 2.4 KB
- references/html-and-canvas.md 2.7 KB
- references/svg-and-vector.md 7.4 KB
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.
- today First seen · 236 lines · 81 tokens per session scan A 065a8dd1d1d8
visual-artifact-design is a skill published in the GitHub repository EverMind-AI/Raven (3,825 stars, last pushed today), licensed Apache-2.0. It adds 81 tokens to every session and 4,580 once invoked, about $0.0004 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 2026-09-12.
Other skills, from other repositories
claude-design
Design one-off HTML artifacts (landing, deck, prototype).
tldraw-offline
Drive and script tldraw offline canvases with an agent.
sketch
Throwaway HTML mockups: 2-3 design variants to compare.
baoyu-infographic
An infographic generator that turns supplied content into a visual summary using different information layouts and visual styles. It supports standard and custom image proportions and multiple languages.
excalidraw
Hand-drawn Excalidraw JSON diagrams (arch, flow, seq).
adversarial-ux-test
Roleplay a hostile user to find and triage UX pain points.