Skip to main content

web-artifacts-builder-codex

Use when Codex needs to create, run, package, revise, or verify a standalone web artifact such as an HTML report, dashboard, microsite, interactive explainer, data story, or browser-delivered deliverable with runtime, screenshots, console capture, and evidence packaging.

来源信息

仓库
dachent/skills
最近来源活动
2026年7月20日 12:04
检测到的 SKILL.md 语言
英语
星标
36
Forks
2

安装方式

默认使用会先检查来源的 Prompt;你也可以切换为直接命令,或下载本地副本。

检查来源文件

决定是否安装前,请先阅读 SKILL.md,以及 SkillsMP 当前展示的配套文件。

文件资源管理器
2 个文件

正在显示 SKILL.md

SKILL.md
来源说明 · 只读预览
name
web-artifacts-builder-codex
description
Use when Codex needs to create, run, package, revise, or verify a standalone web artifact such as an HTML report, dashboard, microsite, interactive explainer, data story, or browser-delivered deliverable with runtime, screenshots, console capture, and evidence packaging.
# Web Artifacts Builder Codex ## Overview Use this skill to build standalone web artifacts that open cleanly, explain themselves through the interface, and ship with enough runtime and visual evidence for another reviewer to trust the result. ## Design Upskill Contribution This skill teaches Codex to treat web artifacts as deliverables, not just files. It improves no-template design by pairing layout and interaction choices with package evidence: screenshot captures, console capture, request failures, PDF export when useful, and a concise evidence bundle. It matters because standalone artifacts often lack the guardrails of an existing app. The artifact must carry its own hierarchy, navigation, source notes, and failure evidence. ## Workflow 1. Define the artifact type: report, dashboard, explainer, portal, tracker, or prototype. 2. Choose the lightest implementation that can be opened and reviewed reliably. 3. Keep the first screen as the actual artifact experience, not a marketing page. 4. Make data/source limitations visible in the artifact when they affect interpretation. 5. Choose the reviewer runtime: direct file open, local server, static export, or a retained dev server. Prefer the simplest path that matches the real delivery constraint. 6. Use stable file names and relative asset paths so the artifact can move as a folder; verify generated files and required environment values from that same runtime. 7. Capture screenshots, console output, request failures, and optional PDF output from the reviewer runtime. 8. Package the artifact with evidence, stable filenames, rerun instructions, and a short verification note. Do not commit generated evidence unless the project explicitly requests it. ## Shared Visual Runtime Use `.shared\visual-runtime` for artifact verification: - `capture_page.mjs` to render local HTML or a localhost URL and capture screenshots. - `export_pdf.mjs` when the artifact needs print or handoff fidelity. - `visual_lint.mjs` for obvious visual and console defects. - `image_bounds.py` and `make_contact_sheet.py` for asset review and contact sheets. No Office COM is required. If a standalone web artifact is later embedded into an Office file, that embedding/export step belongs to the relevant Office COM skill. ## Verification Before completion: - open or serve the standalone web artifact exactly as a reviewer would; - verify relative assets load; - run visual lint; - preserve screenshot, manifest, console, request-failure, and optional PDF evidence; - confirm the evidence bundle can be understood without hidden conversation context and explain how to rerun it. ## Example Prompts ### HTML Memo And Dashboard ```text Use $web-artifacts-builder-codex and $frontend-design-codex. Create a polished HTML memo and dashboard from the materials in: [PASTE FOLDER OR FILE PATHS] Objective: Build an executive-ready HTML artifact that explains [TOPIC / DECISION / ANALYSIS] and includes both a written memo and an interactive dashboard. Requirements: - Do not use a template. Derive the visual system from the subject matter and source content. - Include an executive summary, key findings, supporting analysis, risks/limitations, and recommended next steps. - Include dashboard sections with KPI cards, charts, tables, filters or segmented views where useful. - Make it responsive for desktop and mobile. - Use the shared visual runtime for browser screenshots, PDF export, image bounds checks, and visual linting. - Run accessibility and visual QA checks. - Provide evidence artifacts: screenshots, PDF export, visual lint output, and a short QA note. - Do not use Office COM unless exporting into an Office file becomes explicitly necessary. - Save all outputs under: [OUTPUT FOLDER] ``` ## Common Mistakes - Shipping an artifact that only works from a dev server when the user needed a portable folder. - Hiding caveats in conversation instead of the artifact or evidence note. - Forgetting mobile or print views. - Treating package evidence as optional.
在 GitHub 查看