Przejdź do treści
Biblioteka skilli
Dokumenty i biuro · Pisanie

html-effectiveness

/html-effectiveness

Tworzy pojedynczy samodzielny plik HTML zamiast ściany markdownu: raport, dashboard, tablicę triage, dokument projektowy, review, postmortem lub talię promptów z układem i interakcją.

Licencja
MIT
Ostatnia zmiana repozytorium
12.10.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
25
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
1
Przestarzałość

Brak sygnałów. Repozytorium zmienione 12.10.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Tworzy pojedynczy samodzielny plik HTML zamiast ściany markdownu: raport, dashboard, tablicę triage, dokument projektowy, review, postmortem lub talię promptów z układem i interakcją.

Wyzwalacze z opisu

html instead of markdownmake me a dashboard
Oryginalny opis (EN)

Produce a single self-contained .html file instead of a wall of markdown when the user wants a report, dashboard, triage board, design doc, code review, flowchart, slide deck, postmortem, implementation plan, prompt tuner, feature-flag editor, or an agent-spawn deck (HTML carrying ready-to-paste prompts for parallel fan-out) — any artifact that benefits from layout, interaction, drag-drop, side-by-side comparison, or copy-to-clipboard export. Use when the user asks (English OR German, case-insensitive, typo-tolerant, semantic, NOT literal): "html instead of markdown", "make me a dashboard", "h

Jak działa

  1. Canonical source — vendored in-tree
  2. Decision protocol
  3. Trigger principles — read these before matching
  4. Palette choice
  5. Composition (multi-pattern output)
  6. When to skip this skill
  7. Examples (intent → pattern)
  8. Bundled asset
Sekcje z nagłówków SKILL.md, w oryginale · 87 linii, 1 134 słów
Podgląd SKILL.md
---
name: html-effectiveness
description: 'Produce a single self-contained .html file instead of a wall of markdown when the user wants a report, dashboard, triage board, design doc, code review, flowchart, slide deck, postmortem, implementation plan, prompt tuner, feature-flag editor, or an agent-spawn deck (HTML carrying ready-to-paste prompts for parallel fan-out) — any artifact that benefits from layout, interaction, drag-drop, side-by-side comparison, or copy-to-clipboard export. Use when the user asks (English OR German, case-insensitive, typo-tolerant, semantic, NOT literal): "html instead of markdown", "make me a dashboard", "html report", "triage board", "status report", "wochenbericht", "postmortem", "implementierungsplan", "side by side", "vergleiche x y z", "code review", "slide deck", "präsentation", "feature flag editor", "spawn agents", "agenten spawnen". Fire eagerly when the output would otherwise be a long markdown table or a sortable/filterable list. Match on intent: tickets→triage-board, diff→code-review.'
compatibility: >
  Works in any environment where the agent can write files. The output is a vanilla .html with inline
  CSS/JS — open in a browser, no build, no npm.
---
# html-effectiveness

Based on Thariq Shihipar's [html-effectiveness](https://github.com/ThariqS/html-effectiveness) repo. He works on Claude Code at Anthropic; the philosophy: **agents producing HTML files instead of markdown walls** because HTML carries spatial information and interaction that markdown flattens.

## Canonical source — vendored in-tree

The 20 canonical patterns ship with this plugin as standalone `.html` files under
`${CLAUDE_PLUGIN_ROOT}/skills/html-effectiveness/references/patterns/` (one per `pattern.file`,
e.g. `01-exploration-code-approaches.html` … `20-editor-prompt-tuner.html`). No clone, no network,
no external setup — the plugin carries its own source of truth and works on any machine. They are
the upstream reference (Apache-2.0, see `references/patterns/UPSTREAM-LICENSE`): read them to learn
the pattern, then write a **new** file with the user's data into the user's project area (default:
`~/<task>/<slug>.html` or alongside related artifacts) — never edit the vendored references in place.

## Decision protocol

1. **Load the routing data.** Read `references/patterns.json` from this skill folder. It contains 20 canonical patterns and a `decision_tree` mapping user-intent phrases (English + German + semantic) to `use_pattern` IDs. The `triggering_principles` block at the top tells you HOW to match (fuzzy, multilingual, typo-tolerant, intent-based — not literal).

2. **Match user intent.** Scan the user's request for semantic fit with any entry in `decision_tree`. Match on intent regardless of language, case, or typos. If multiple patterns fit, pick the one whose `key_pattern` field most closely describes the data structure the user has. If still ambiguous, prefer the more interactive pattern (editor > report > diagram).

3. **Study the canonical source.** Open `${CLAUDE_PLUGIN_ROOT}/skills/html-effectiveness/references/patterns/<pattern.file>` and read it end-to-end. Don't paraphrase the pattern — internalize the HTML structure, CSS tokens, JS interaction code, and copy-button mechanism. Reuse the same eyebrow / h1 / sub / sticky-toolbar / column / card markup so the output feels like one family.

4. **Replicate with the user's real data.** Replace Thariq's example data (Acme tickets, Birchline comments, etc.) with the user's actual data. Keep the design tokens (`#FAF9F5` ivory bg, `#D97757` clay accent, `ui-serif, Georgia, serif` + system-ui + SF Mono), layout, and interaction code intact. If the user has no real data yet, ask once for a sample — do not invent fictional company data.

5. **Honor the principles** (see `references/patterns.json:principles`):
   - Self-contained: inline `<style>` and `<script>`, no external dependencies — fully offline, no Google Fonts.
   - Anthropic Coral DNA: design tokens in `references/patterns.json:design_tokens`.
   - Vanilla JS only. No React, no jQuery.
   - Editors end with a `Copy as markdown` (or `Copy diff` / `Copy JSON` / `Copy gh commands`) button that exports the UI state back into something the user can paste into the next prompt.
   - If the artifact contains private data (private repo names, customer info, internal URLs), save it under the user's home (e.g. `~/repo-audit/`, `~/<task>/`) — never commit to a public plugin repo.

6. **Output and open.** Write the `.html` at a sensible location (default: `~/<task>/output.html` or alongside related artifacts). Open it in the user's browser via `open <file>` for immediate inspection. For interactive editors, mention the export-button location so the user knows how to round-trip the UI state.

## Trigger principles — read these before matching

- **Match on intent, not phrase.** The trigger lists are EXAMPLES of intent, not a whitelist. If the user describes a deliverable that would render better with layout + interaction than as markdown prose, fire.
- **Languages: English + German + mixed.** Triggers are listed bilingually. Mixed-language input ("mach mir einen status report"), Austrian/Swiss German variants, and informal contractions all count. Match by semantic intent regardless of language.
- **Typo tolerance.** User typos ("reprot", "staus", "tirage", "flowshart", "implementierungsplan" misspelled) are expected. Match by stem / phonetic similarity.
- **Case-insensitive.** "HTML REPORT", "html report", "Html Report" are the same.
- **Ambiguous user intent.** If the user says "make me a thing for X", look at WHAT X is. A list of tickets → triage-board. A diff → code-review. A timeline → incident-report. Let the data shape decide the pattern.
- **Be pushy.** Default to firing when the output would otherwise be a long markdown table, several sections of prose with categories, or a list the user will sort/filter/diff. Markdown for a paragraph; HTML for anything you'll come back to.

## Palette choice

`patterns.json:palettes` defines two:

- **`coral`** (default) — `#FAF9F5`/`#D97757`/`#141413`/`#788C5D` + `ui-serif, Georgia, serif`. Use for reports, plans, design docs, status updates, research explainers, editor UIs. Anything reading-room or knowledge-work.
- **`github_dark`** — `#0d1117`/`#58a6ff`/`#3fb950`/`#f85149`. Use for codebase audits, CI/CD dashboards, fan-out agent decks, terminal-adjacent tooling. Anything that lives next to a terminal session.

Pattern `21-agent-spawn-deck` defaults to `github_dark`. All others default to `coral`. The user can override per request.

## Composition (multi-pattern output)

If a task naturally splits across multiple deliverables (e.g., an implementation-plan AND a spawn-deck for executing the plan), write **multiple linked .htmls** and reference each other via `<a href>` at the bottom. Example: `repo-cleanup-plan.html` (pattern 16, plan) + `repo-cleanup-spawn-deck.html` (pattern 21, parallel scanners) + `repo-cleanup-triage.html` (pattern 18, result triage). Each links the next.

If a single pattern needs sub-work that parallelizes (audit 200 repos, scan 30 services), default to `21-agent-spawn-deck` instead of one giant report — the deck carries N ready-to-paste prompts, the user fan-outs, the lead synthesizes.

## When to skip this skill

- The user explicitly asked for markdown / a commit message / a code change in a `.py` / `.cs` / etc. file.
- The output is genuinely a single short paragraph or single code snippet.
- The user wants prose they will paste into another system that doesn't render HTML (e.g., a Slack message body, a commit body, a CHANGELOG entry).

## Examples (intent → pattern)

| User says | Pattern fires |
|---|---|
| "triage diese 200 dead repos" | `18-editor-triage-board` |
| "make me a weekly status of what shipped" | `11-status-report` |
| "wochenbericht für Sprint 14" | `11-status-report` |
| "zeig mir die generated C# vs YAML side by side" | `01-exploration-code-approaches` |
| "explain how auth flows through this repo" | `14-research-feature-explainer` (or `04-code-understanding` if structural) |
| "tune my SKILL.md description against these 10 prompts" | `20-editor-prompt-tuner` |
| "incident bei deploy gestern, mach postmortem" | `12-incident-report` |
| "implementierungsplan für ANcpLua.NET.Sdk v3 mit Risks" | `16-implementation-plan` |
| "annotated diff für PR #145" | `03-code-review-pr` |
| "präsi für Bachelor-Verteidigung mit Pfeiltasten" | `09-slide-deck` |
| "feature flags editor für flags.production.json" | `19-editor-feature-flags` |
| "concept explainer für consistent hashing, interaktiv" | `15-research-concept-explainer` |
| "fan out 4 agents to scan my 349 repos" | `21-agent-spawn-deck` (github_dark palette) |
| "gib mir das deck mit prompts zum parallel ausführen für den shadow-ai audit" | `21-agent-spawn-deck` |
| "implementation plan PLUS spawn-deck zum ausführen" | `16-implementation-plan` + `21-agent-spawn-deck` (linked) |

## Bundled asset

`assets/otel-semconv-demo.html` is a concrete proof-of-concept built on top of `01-exploration-code-approaches` pattern, showing OTel semantic-convention YAML → generated C# diff with breaking-change filter. Open it to see the pattern applied to Alexander's actual problem domain.
Pliki w paczce (25)
  • SKILL.md 9 KB
  • assets/agent-spawn-deck-demo.html 19 KB
  • assets/otel-semconv-demo.html 17 KB
  • references/patterns.json 25 KB
  • references/patterns/01-exploration-code-approaches.html 14 KB
  • references/patterns/02-exploration-visual-designs.html 14 KB
  • references/patterns/03-code-review-pr.html 26 KB
  • references/patterns/04-code-understanding.html 18 KB
  • references/patterns/05-design-system.html 18 KB
  • references/patterns/06-component-variants.html 16 KB
  • references/patterns/07-prototype-animation.html 12 KB
  • references/patterns/08-prototype-interaction.html 11 KB
  • references/patterns/09-slide-deck.html 16 KB
  • references/patterns/10-svg-illustrations.html 18 KB
  • references/patterns/11-status-report.html 16 KB
  • references/patterns/12-incident-report.html 15 KB
  • references/patterns/13-flowchart-diagram.html 15 KB
  • references/patterns/14-research-feature-explainer.html 12 KB
  • references/patterns/15-research-concept-explainer.html 13 KB
  • references/patterns/16-implementation-plan.html 28 KB
  • references/patterns/17-pr-writeup.html 24 KB
  • references/patterns/18-editor-triage-board.html 18 KB
  • references/patterns/19-editor-feature-flags.html 19 KB
  • references/patterns/20-editor-prompt-tuner.html 21 KB
  • references/patterns/UPSTREAM-LICENSE 11 KB

Wersje

W bazie nie ma innych skilli o nazwie html-effectiveness.

Migawka bazy z 13.11.2026 · wersja paczki a1533a013912 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100html-artifacts
    Tworzy samodzielny artefakt HTML zamiast dokumentu Markdown, gdy treść wymaga układu, kolorów, diagramów lub interaktywności, na przykład plan, specyfikacja, raport, porównanie albo dashboard.
  • 100html-output
    Zapisuje plany, specyfikacje, raporty i przeglądy kodu jako jeden samodzielny plik HTML zamiast długiego Markdownu.
  • 90md2deck
    Tworzy prezentację z markdownu w jednym pliku HTML i sprawdza każdy slajd pod kątem tytułu, czytelności i rozmiaru tekstu.
  • 100commentable-html
    Zamienia samodzielny raport, plan, dashboard lub dokument HTML w powierzchnię do komentowania: recenzenci zaznaczają fragmenty, dodają komentarze w tekście i eksportują wątek do agenta jako pakiet maszynowy.
  • 100html-brief-artifact
    Tworzy samodzielny brief HTML dla planów, statusów, podsumowań PR, notatek o incydentach i wyjaśnień badań, jako dopracowany dokument zamiast markdownu, slajdów czy pełnego frontendu.
  • 100render-html
    Zamienia artefakt Markdown lub JSON z ARIS na jeden plik HTML do czytania przez człowieka, z przejrzystym układem.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.