{"id":50369382,"url":"https://github.com/hyun2xyz/iiki","last_synced_at":"2026-05-30T06:04:17.037Z","repository":{"id":360622862,"uuid":"1250965333","full_name":"hyun2xyz/iiki","owner":"hyun2xyz","description":"Wikibird research-to-HTML Agent Skill for Korean wiki-style explainers","archived":false,"fork":false,"pushed_at":"2026-05-29T10:00:24.000Z","size":1447,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-29T12:05:28.649Z","etag":null,"topics":["agent-skills","claude-code","codex","gemini-cli","research-agent","skill-md","static-html"],"latest_commit_sha":null,"homepage":null,"language":"HTML","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/hyun2xyz.png","metadata":{"files":{"readme":"README.en.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-27T05:57:12.000Z","updated_at":"2026-05-29T10:00:39.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/hyun2xyz/iiki","commit_stats":null,"previous_names":["hyun2xyz/wiki_bird","hyun2xyz/iiki"],"tags_count":3,"template":false,"template_full_name":null,"purl":"pkg:github/hyun2xyz/iiki","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hyun2xyz%2Fiiki","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hyun2xyz%2Fiiki/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hyun2xyz%2Fiiki/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hyun2xyz%2Fiiki/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/hyun2xyz","download_url":"https://codeload.github.com/hyun2xyz/iiki/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hyun2xyz%2Fiiki/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33681815,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-05-30T02:00:06.278Z","response_time":92,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-skills","claude-code","codex","gemini-cli","research-agent","skill-md","static-html"],"created_at":"2026-05-30T06:04:16.259Z","updated_at":"2026-05-30T06:04:17.032Z","avatar_url":"https://github.com/hyun2xyz.png","language":"HTML","funding_links":[],"categories":[],"sub_categories":[],"readme":"# iiki\n\n**2026-05-29 · v0.2.0 · IYO Wiki text and Typst booklet beta (ദ്ദി ᴖ⩊ᴖ )**\n\n```text\nHello, I'm Hyun.\n\nI study editorial design in Korea, research different AI-powered pipelines,\nand share them with other people.\nI love the sharing philosophy of the World Wide Web, HTML, humor, and writing.\n\n***\n\nOn May 9, Thariq Shihipar posted on X about how the Claude Code team uses HTML.\nThe point was that, rather than asking AI for something and receiving the result\nas plain text or Markdown, an HTML prototype is often easier to read and judge.\nIt is a little like how we usually do not notice it, but through graphic design\nwe can take in a lot of information without reading every bit of text.\nThe way to use it is simple: at the end of any conversation, say,\n\"make this into HTML.\"\n\nBut the HTML UI made this way can be a little uncomfortable.\nMore importantly, it is very ugly.\n\n***\n\nLet me introduce my friend iiki.\n\nIt is a bird we raise in our IYO Wiki.\nIts main service is that it adds 짹 to the end of every sentence, which is cute.\n\nAs a side service, when you ask it to research a topic, it organizes the content\nin a wiki style and attaches image references using Chrome captures.\nEverything is saved locally as MD and HTML, so if you open index.html you can see\nall the previous records at once. Of course, you can also host that file.\n\nThen, use it well.\n```\n\n[한국어 README](README.md)\n\n![iiki mini wiki mascot](docs/assets/miniwiki.jpg)\n\n\u003e **Hello. I can organize anything for you, 짹**\n\niiki is a research agent/skill that turns a topic into a readable wiki-style HTML document.\n\nGive it something to research, and it checks current sources, keeps citations, captures useful screenshots, and writes a compact wiki page with a table of contents. When needed, it can also create Typst/PDF output.\n\nThe goal is not to translate source pages line by line. iiki checks the material, then rewrites it into a readable explanation: terms are unpacked before they are used, context is added where it changes the meaning, and tables or images are pulled out when they help the reader verify what is being said. The voice should stay factual, not chatty; when judgment is needed, the evidence sits next to it.\n\nThe root `index.html` is a local-only index of generated research pages under `outputs/`. For public GitHub Pages documentation, use `docs/index.html`.\n\n## What It Does\n\n- Topic research: start from official, primary, and recent sources whenever possible.\n- Source-backed synthesis: keep evidence near important claims and mark uncertainty instead of smoothing it over.\n- Korean wiki-style writing by default: rewrite the flow for Korean readers instead of translating source sentences one by one.\n- Static HTML page with a fixed table of contents: make long research easy to scan and revisit.\n- Chrome screenshot capture: keep visual proof for people, services, papers, official pages, and interfaces that need to be seen.\n- `iyo-wiki.txt` export for IYO Wiki copy-paste: turn the HTML result into wiki-friendly text.\n- `research.md` and `qa.md` logs: record what was checked, what passed, and what remains limited.\n- Optional Typst/PDF booklet output: create a separate print-friendly document version from the same research.\n\nDefault output structure:\n\n```text\noutputs/YYYY-MM-DD-topic-slug/\n├── index.html\n├── iyo-wiki.txt\n├── research.md\n├── qa.md\n├── assets/\n│   └── screenshots/\n│       └── 01-source-or-page.png\n└── typst/\n    ├── iiki-booklet.typ\n    ├── main.typ\n    └── index.pdf\n```\n\nEach output folder has its own `index.html`, which can be opened directly in a browser. New research outputs should be created in a specified folder or under this repo's `outputs/`, not on the Desktop by default.\n\n## Screenshots\n\nThe root `index.html` is a simple local list.\n\n![iiki local index screenshot](docs/assets/screenshots/local-index.png)\n\nEach research record lives in `outputs/YYYY-MM-DD-topic-slug/`. Open that folder's `index.html` locally to read the generated wiki page.\niiki documents put the left table of contents, body text, images, and sources together on one page.\n\n![iiki wiki page screenshot](docs/assets/screenshots/iiki-manual-page.png)\n\n## Install And Activate\n\nThe recommended way to use iiki is through CLI-based agents. If you want research output saved as files, Chrome screenshots captured, and checks like `node scripts/build-index.js` run reliably, use Codex CLI or Claude Code.\n\n### 1. Install For Codex CLI\n\n```sh\ngit clone https://github.com/hyun2xyz/iiki.git\ncd iiki\nscripts/install-skill.sh codex\n```\n\nThe current Codex user skill location is `$HOME/.agents/skills/iiki-research-html/`. For repo scope, place the skill at `.agents/skills/iiki-research-html/`.\n\nFor older Codex setups, install into the legacy location:\n\n```sh\nscripts/install-skill.sh codex-legacy\n```\n\nInvoke it in Codex like this:\n\n```text\nUse $iiki-research-html\n```\n\n### 2. Install For Claude Code\n\n```sh\ngit clone https://github.com/hyun2xyz/iiki.git\ncd iiki\nscripts/install-skill.sh claude\n```\n\nFor Claude Code, keep the skill folder in `~/.claude/skills/iiki-research-html/` or project `.claude/skills/iiki-research-html/`.\n\nInvoke it in Claude Code like this:\n\n```text\nUse iiki-research-html\n```\n\nThe install script overwrites any existing iiki skill folder at the target location.\n\n### 3. Set Research Depth\n\nAfter first installing or teaching iiki, ask the user to choose a default research depth:\n\n```text\nWould you like to set a default research depth? Choose 하, 중, or 상.\n하: quick overview, 중: normal iiki document, 상: deep study research.\n```\n\n### 4. Manual PowerShell Install For Codex CLI\n\n```powershell\ngit clone https://github.com/hyun2xyz/iiki.git\ncd iiki\n\n$dest = Join-Path $HOME \".agents\\skills\\iiki-research-html\"\nNew-Item -ItemType Directory -Force (Split-Path $dest) | Out-Null\nRemove-Item -Recurse -Force $dest -ErrorAction SilentlyContinue\nCopy-Item -Recurse \".\\skills\\iiki-research-html\" $dest\n```\n\n### 5. Check Installation\n\n1. Restart Codex or Claude Code.\n2. Invoke `$iiki-research-html` or `Use $iiki-research-html`.\n3. On first use, choose a default research depth: `하`, `중`, or `상`.\n4. If the skill does not appear, check that `SKILL.md` is directly inside the installed skill folder.\n\nFor Codex, the expected user-scope structure is:\n\n```text\n$HOME/.agents/skills/iiki-research-html/SKILL.md\n```\n\n### 6. One-Minute App Trial\n\nFor ChatGPT, Claude, and Gemini apps, this README does not provide install or activation steps. Use the prompt below to make the model follow iiki for the current conversation only.\n\n```text\nUse the iiki method based on https://github.com/hyun2xyz/iiki.\nRead README.md and skills/iiki-research-html/SKILL.md, then follow them.\n\nTopic: Why HTML prototypes are useful as AI research outputs\nResearch depth: 중\nOutput:\n- index.html body structure\n- research.md\n- qa.md\n- screenshot candidates\n- source list\n\nWrite it as a Korean wiki-style explanation.\n```\n\nThis is only for previewing the structure. App chats may not reliably create local files, save Chrome screenshots, or refresh the local index. Use a CLI environment for real iiki outputs.\n\n### Troubleshooting\n\n- If an app returns a normal chat answer: it probably treated the GitHub URL as \"read this for the current conversation.\" Use the one-minute trial prompt above and explicitly name `README.md`, `SKILL.md`, and the expected output files.\n- If the HTML does not look like iiki: the app may be inventing its own style because it cannot write and verify the repo files. Use a CLI environment for real HTML output.\n- If Codex cannot see the skill: check `$HOME/.agents/skills/iiki-research-html/SKILL.md`, then restart Codex.\n- If you are on an older Codex setup: run `scripts/install-skill.sh codex-legacy` to install into `~/.codex/skills/`.\n\nOfficial references:\n\n- OpenAI Developers, [Codex Skills](https://developers.openai.com/codex/skills)\n- Anthropic Docs, [Agent Skills - Claude Code](https://code.claude.com/docs/en/skills)\n\n## Quick Start\n\nAsk an agent that supports skills:\n\n```text\nUse $iiki-research-html.\nResearch the book A New Program for Graphic Design and turn it into outputs/YYYY-MM-DD-graphic-design-course/index.html.\nCheck current sources, include Chrome screenshots, and keep research.md and qa.md.\n```\n\nYou can also call it more casually:\n\n```text\nUse iiki to research the Typst template ecosystem and make a wiki-style HTML page.\n```\n\nOnce the agent has learned the skill, this can be even shorter:\n\n```text\nUse iiki for a Typst research note.\n```\n\nWhen iiki is active, conversational replies should end with `짹`.\n\n```text\nI'll organize it, 짹\nDone, 짹\n```\n\nResearch depth can be set to `하`, `중`, or `상`. If no depth is specified, the default is `중`.\n\n```text\nUse $iiki-research-html to research Andrej Karpathy.\nResearch depth: 상\nInclude basic profile information, a public profile/face reference, and representative work pages.\n```\n\n| Depth | Sources | Images | Use |\n| --- | --- | --- | --- |\n| `하` | 3-5 sources | Essential captures only | Quick overview |\n| `중` | 6-10 sources | As needed for major sections | Normal iiki document |\n| `상` | 10+ sources when available | Add enough for understanding and verification | Deep study document |\n\n## Chrome Screenshots\n\niiki expects at least one Chrome-captured image in research HTML outputs.\n\nGuidelines:\n\n- Capture an official doc page, product page, paper/source page, or the final rendered HTML.\n- For people, prefer official profiles, personal sites, university/company pages, talks, or other reputable public pages.\n- Do not screenshot every keyword. Use images that help the reader understand the topic. There is no fixed maximum; add more when they clarify the research.\n- Save images under `assets/screenshots/`.\n- Embed screenshots with `\u003cfigure class=\"screenshot\"\u003e`.\n- Log URL, capture date, viewport, and purpose in `research.md`.\n- Do not capture private dashboards, personal data, paywalled text, passwords, cookies, or logged-in account screens.\n\nIf the agent has no native Chrome tool, use the helper:\n\n```sh\nskills/iiki-research-html/scripts/capture-chrome-screenshot.sh \\\n  \"https://example.com\" \\\n  outputs/YYYY-MM-DD-topic-slug/assets/screenshots/01-example.png\n```\n\n## Viewing Results\n\nOpen a research page directly:\n\n```sh\nopen outputs/YYYY-MM-DD-topic-slug/index.html\n```\n\nResearch records are checked by opening each output folder's `index.html` locally.\n\nRefresh and open the local root index:\n\n```sh\nnode scripts/build-index.js\nopen index.html\n```\n\nRun `node scripts/build-index.js` after creating or editing an output so the root `index.html` stays current.\n\n## IYO Wiki Copy-Paste Text\n\nIf you need text that can be pasted into IYO Wiki, run this after the final HTML is finished. This is not a second research result; it is an intermediate form for moving the already organized page into a wiki editor.\n\n```sh\nnode scripts/export-iyo-wiki-text.js outputs/YYYY-MM-DD-topic-slug\n```\n\nThis creates `iyo-wiki.txt` in the same output folder. It converts headings, paragraphs, links, lists, tables, quotes, and image captions into IYO Wiki-friendly text. The export keeps as much as possible editable, so the pasted page can still be revised inside the wiki.\n\nImages are exported as `{{iyoimage ...}}` placeholders. After uploading files to IYO Wiki, adjust the `file=\"...\"` paths as needed.\n\n## Typst/PDF Booklet\n\nFor a paper-like booklet PDF, run this after the HTML is finished. HTML is the screen-reading version; Typst/PDF is the saved, shared, or printed version. The idea is to derive a document edition from the same research, not to write the research twice.\n\n```sh\nnode scripts/build-typst-booklet.js outputs/YYYY-MM-DD-topic-slug\n```\n\nThis creates `typst/main.typ`, `typst/iiki-booklet.typ`, and `typst/index.pdf`. The default layout is close to an A5 booklet, with calmer margins and line lengths for longer reading. If the Typst CLI is unavailable, the `.typ` files can still be generated and compiled later.\n\nTo generate Typst source without compiling PDF:\n\n```sh\nnode scripts/build-typst-booklet.js outputs/YYYY-MM-DD-topic-slug --no-compile\n```\n\n## Main Files\n\n- `skills/iiki-research-html/SKILL.md` - installable skill body\n- `skills/iiki-research-html/references/` - research, HTML, and Typst rules\n- `skills/iiki-research-html/scripts/capture-chrome-screenshot.sh` - Chrome screenshot helper\n- `scripts/export-iyo-wiki-text.js` - IYO Wiki copy-paste text export\n- `scripts/build-typst-booklet.js` - Typst/PDF booklet export\n- `.claude/agents/iiki-research-html-builder.md` - Claude Code project agent\n- `index.html` - local-only research output index\n- `docs/index.html` - public iiki user guide for GitHub Pages\n- `docs/research-html-workflow.md` - workflow docs\n- `docs/wiki-html-style-rules.md` - HTML style rules\n- `docs/distribution-plan.md` - Codex, Claude, Gemini, and Cursor distribution notes\n- `outputs/` - generated research page examples\n\n## Writing Tone\n\nDefault writing style is Korean wiki-style explanation.\n\n- Avoid AI-ish filler.\n- Use short paragraphs.\n- Separate sourced facts from interpretation.\n- Use exact dates for current information.\n- Start with the flow, then add tables and screenshots.\n\n## License\n\nMIT. The skill folder also includes its own `LICENSE`, so it can be distributed separately.\n\n## Then, Fighting\n\nFighting\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhyun2xyz%2Fiiki","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fhyun2xyz%2Fiiki","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhyun2xyz%2Fiiki/lists"}