just-html

Publish, edit, and collaborate on HTML documents at justhtml.sh. Covers agent self-onboarding (the auth.md emailed-code flow to get an API key) and the full doc / comment / reaction API with curl examples. Use when asked to publish or host an HTML document, spec, doc, outline, or proposal; to share one with a person or their agent; or to comment on / edit a justhtml.sh document.

Ir para a instalação

Repositório: kernel/just-html
Última atividade na origem: 27 de julho de 2026 às 17:22
Idioma detectado do SKILL.md: inglês
Estrelas: 20
Forks: 0

Opções de instalação

Por padrão, está selecionado o prompt que primeiro revisa a origem. Você pode mudar para um comando direto ou baixar uma cópia local.

Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.

Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.

npx skills add https://github.com/kernel/just-html --skill just-html

O comando permanece em uma só linha. Role horizontalmente para revisá-lo antes de copiar.

Prefere uma cópia local? Baixe os arquivos disponíveis atualmente no SkillsMP.

Revise os arquivos de origem

Leia o SKILL.md e os arquivos complementares exibidos pelo SkillsMP antes de decidir se vai instalar.

SKILL.md

Instruções da origem · Visualização somente leitura

Voltar à pesquisa Voltar à instalação

stripe-projects-cli

Use the Stripe Projects CLI in this repository to manage deploying and access to third party services.

2026-06-12

Ver todas deste repositório

name	just-html
description	Publish, edit, and collaborate on HTML documents at justhtml.sh. Covers agent self-onboarding (the auth.md emailed-code flow to get an API key) and the full doc / comment / reaction API with curl examples. Use when asked to publish or host an HTML document, spec, doc, outline, or proposal; to share one with a person or their agent; or to comment on / edit a justhtml.sh document.

justhtml.sh

An agent-first minimal HTML document host. Your agent self-onboards, gets a long-lived API key, and publishes HTML to stable URLs like https://justhtml.sh/d/fierce-tiger-12345. Docs are private by default, shareable via a view token, and optionally public. Humans and their agents collaborate on the same documents.

Everything here is reachable with curl. No SDK required. JSON in, JSON out; Authorization: Bearer jh_live_... on the API.

Authentication

Full prose protocol (auth.md service_auth flow): https://justhtml.sh/auth.md Machine-readable discovery: https://justhtml.sh/.well-known/oauth-protected-resource https://justhtml.sh/.well-known/oauth-authorization-server

Short version: you can't self-issue a key. There is exactly ONE flow. Register with the human's email; we email them a 6-digit code; they read it back to you; you submit it and poll for the key. Steps:

1. Start registration (no account is created yet). We email the human a

6-digit code. The user_code is NOT in the response.

curl -s https://justhtml.sh/agent/identity
-H 'Content-Type: application/json'
-d '{"type":"service_auth","login_hint":"you@example.com"}'

-> { claim_token, claim: { complete_url, expires_in, interval } }

2. Tell the human: "check your email and tell me the 6-digit code from the

","title":"Demo","public":false}' ","newText":""}],"base_version":1}'

just-html

Informações da origem

Opções de instalação

Revise os arquivos de origem

Mais deste repositório

Mais deste repositório

justhtml.sh

Authentication

1. Start registration (no account is created yet). We email the human a

6-digit code. The user_code is NOT in the response.

-> { claim_token, claim: { complete_url, expires_in, interval } }

2. Tell the human: "check your email and tell me the 6-digit code from the

justhtml.sh

Authentication

1. Start registration (no account is created yet). We email the human a

6-digit code. The user_code is NOT in the response.

-> { claim_token, claim: { complete_url, expires_in, interval } }

2. Tell the human: "check your email and tell me the 6-digit code from the

justhtml.sh message." When they give it to you, submit it:

-> 200 {status:"claimed"}; wrong code -> 401 (5 tries, then 410 code_dead)

3. Poll for the credential (every claim.interval = 5s). Form-encoded.

-> authorization_pending until they finish; then { access_token: "jh_live_..." }

Code expired before they read it back? Re-mint (24h registration window) —

emails a fresh code and invalidates the old one:

Revoke a key (RFC 7009, idempotent):

API (base: https://justhtml.sh/api/v1)

Hi

-> 201 { slug, url, view_token, version, public, ... }

Private doc share link: ?viewtoken=<view_token>

scope=owned (default): docs you own. scope=shared: docs granted to your

email or your email's domain (excludes docs you own). scope=all: both.

Each item: { slug, url, title, access, version, public, comment_count,

created_at, updated_at }. access is owner|editor|commenter|

viewer (an explicit email grant beats a domain grant).

comment_count is live comments+replies on the doc. Owned items

also carry view_token; shared items do not.

The signed-in web equivalent (owned + shared sections) is https://justhtml.sh/docs

Hi

Hello

Always send base_version. Mismatch -> 409 with current_version. Ambiguous /

no-match / overlapping edits -> 422 naming the failing edit (retry with more

context).

This endpoint also accepts a signed-in session, which is how the viewer's

inline edit mode saves: an owner or editor grantee opens /d/:slug, clicks the

pencil, and types on the page. It patches TEXT only (typos, rewording) and

posts the same edits you would; structural changes stay an agent job.

-> 201 { slug, grant }

Domain grants (e.g. {"domain":"co.com"}) work too; consumer providers

(gmail.com, ...) are rejected -> use public or the view token instead.

A teammate's agent registers via auth.md with that email and the grant

authorizes their edits.

Email grants send the grantee a share-notification email with ONE link that

logs them in (no account needed) and lands them on /d/:slug — a 7-day,

single-use login link. The email also tells them how to register an agent

via auth.md to edit. Pass {"notify":false} to suppress the email (e.g. you'll

share the link yourself). Domain grants NEVER email (we don't notify a whole

company). Notification sends count against the per-recipient email caps.

Comments & reactions

-> 201 { comment: { id, author, body, anchor, orphaned, resolved, ... } }

-> { total, can_comment, can_react, threads:[ { id, author, body, anchor,

group:"anchored"|"doc"|"orphaned", resolved, orphaned, reactions:[...],

replies:[...] } ] } # anchored threads in document order, then

doc-level, then orphaned. Resolved threads carry resolved:true.

Reply: POST /comments with {"body":"+1","parent_id": }

Edit body (author only): PATCH /comments/:id {"body":"..."}

Resolve/unresolve (anyone who can comment): PATCH /comments/:id {"resolved":true}

Delete (author own, owner any; soft): DELETE /comments/:id

A reaction targets exactly ONE of: a comment (comment_id), a text span

(anchor — same W3C shape as a comment anchor), or the whole doc (neither).

React on the doc:

React on a comment: add {"comment_id":42}.

React on a QUOTED SPAN (anchored reaction — an agent "highlights" by quoting):

-> 201 { reaction: { id, emoji, author, comment_id, anchor, anchored_version,

orphaned, created_at } }

(or 200 { toggled:true, removed:true } if the same reaction existed)

Supplying BOTH comment_id AND anchor is a 400 (the target is mutually exclusive).

Remove a reaction: DELETE /docs/:slug/reactions/:id (your own), or re-POST to toggle.

Reactions are unique per (target, author, emoji) — for spans, "target" is the

anchor signature, so the SAME emoji on two different spans are two reactions.

Anchored reactions re-anchor on every doc edit exactly like comments (move,