@hasna/skills 0.1.60 → 0.1.62

package/skills/audio-extract/SKILL.md

@@ -2,7 +2,7 @@
 name: audio-extract
 version: 0.1.0
 description: Extract audio tracks from video files using FFmpeg with format conversion options
-author: skills.md
+author: Hasna
 category: Media Processing
 tags:
   - audio

package/skills/blog-article/SKILL.md

@@ -1,6 +1,7 @@
 ---
 name: blog-article
-description: Generate one or more SEO-ready blog article packages as hosted premium artifacts.
+description: Generate one or more SEO-ready blog article packages as hosted artifacts.
+kind: instruction
 ---
 
 # Blog Article
@@ -9,32 +10,28 @@ Generate blog articles as downloadable Markdown, HTML, JSON, and manifest artifa
 
 ## Requirements
 
-- Authenticate with `skills auth login`.
-- This premium skill runs through the hosted Skills runtime; local installs expose only skill metadata and instructions.
+None. This is an instruction skill: it is prose an agent follows, so it needs no
+credentials, no network access, and no local runtime.
 
 ## Usage
 
-```bash
-skills quote create-blog-article --count 8 --topic "SaaS onboarding"
-skills run create-blog-article --count 8 --topic "SaaS onboarding" --audience "founders" --length long --seo
-skills run create-blog-article -- --topic "Product-led growth" --articles 12
-```
+Ask an agent for the deliverables below and give it the inputs. Reading this file
+IS the invocation; `skills run` refuses instruction skills on purpose.
 
-## Options
+## Inputs
 
-| Option | Description | Default |
+| Input | Description | Default |
 |--------|-------------|---------|
-| `--topic <text>` | Topic or theme. Positional text also works. | required |
-| `--audience <text>` | Intended reader or buyer persona. | |
-| `--count <number>` | Number of articles to generate, 1-12. | 1 |
-| `--articles <number>` | Alias for `--count`. | |
-| `--tone <tone>` | `professional`, `casual`, `technical`, or `friendly`. | professional |
-| `--length <length>` | `short`, `medium`, or `long`. | medium |
-| `--seo` | Include SEO metadata and keyword-oriented structure. | false |
-| `--outline <text>` | Optional outline, angles, or required sections. | |
-| `--output <dir>` | Output directory for direct local execution. Hosted runs use the run export directory. | run export dir |
-
-## Outputs
+| `topic` | Topic or theme. Positional text also works. | required |
+| `audience` | Intended reader or buyer persona. | |
+| `count` | Number of articles to generate, 1-12. | 1 |
+| `articles` | Alias for `--count`. | |
+| `tone` | `professional`, `casual`, `technical`, or `friendly`. | professional |
+| `length` | `short`, `medium`, or `long`. | medium |
+| `seo` | Include SEO metadata and keyword-oriented structure. | false |
+| `outline` | Optional outline, angles, or required sections. | |
+
+## Deliverables
 
 - `manifest.json` using the `skills.blogArticle.outputs.v1` contract
 - For a single article: `article.md`, `article.html`, `article.json`
@@ -42,4 +39,17 @@ skills run create-blog-article -- --topic "Product-led growth" --articles 12
 
 `manifest.json` includes the run id, prompt, sanitized input options, requested tone/length/SEO settings, receipt cost, and an `articles` array. Each article entry includes `title`, `slug`, `summary`, `keywords`, optional `featuredImage`, word count, reading time, and relative file paths.
 
-After submitting a hosted run, poll with `skills runs status <run-id>` and download outputs with `skills exports download <run-id>`.
+## Method
+
+1. Confirm topic, audience and search intent before writing. If the topic is
+   ambiguous, state the interpretation you chose.
+2. Outline first: H2/H3 skeleton with one sentence of intent per section. Get the
+   shape right before producing prose.
+3. Write the draft to the requested length and tone. Lead with the answer; do not
+   warm up for three paragraphs.
+4. Support claims with specifics — numbers, examples, named tools. Remove any
+   sentence that would survive unchanged in an article about something else.
+5. If SEO is requested, add title, meta description, slug and internal-link
+   suggestions as a separate block, never by stuffing the body.
+6. Finish with a self-edit pass: cut hedging, merge duplicate points, verify every
+   factual claim you cannot source is marked as an assumption.

package/skills/blog-article/package.json

@@ -5,7 +5,6 @@
   "description": "Hosted premium blog article artifact generator",
   "type": "module",
   "skills": {
-    "runtime": "hosted",
-    "source": "remote"
+    "kind": "instruction"
   }
 }

package/skills/brand-assets/SKILL.md

@@ -1,57 +1,118 @@
 ---
 name: brand-assets
-description: Fetch official brand assets from a website or brand name, then return logos, palette, typography, source metadata, and a clean asset manifest. Use when a user asks to find, extract, download, package, or normalize a brand's logo and visual identity assets.
+description: Extract brand assets from a website you name — icons, social preview images, logo images, inline SVG marks, theme colors, CSS color tokens, and font stacks — downloaded into a folder with full source provenance. Use when a user asks to find, extract, download, or package a site's logo and visual identity.
 ---
 
 # Brand Assets
 
-Discover official brand identity material for a company, product, or website and package it into a clean handoff folder.
+Point this at a URL. It fetches that page over plain HTTP, parses the markup and
+the stylesheets it links, and pulls out the brand's icons, social images, logo
+images, colors, and typography. Everything downloaded is recorded against the
+exact URL it came from.
+
+The only hosts contacted are the one you name and whatever that page links its
+own assets from. There is no search API, no AI service, and no key of any kind.
 
 ## Requirements
 
-- Hosted premium execution requires `SKILLS_API_KEY`.
-- Provider keys stay server-side in the hosted runtime.
-- Provide either a brand name, a website URL, or both. A URL is preferred when the brand name is ambiguous.
+- Bun (the skill runs as `bun run src/index.ts`).
+- Outbound HTTP access to the site you pass. Nothing else.
+- No API keys and no environment variables.
+- One runtime dependency: `node-html-parser`. Run `bun install` in this skill
+  directory. If it is missing you get
+  `Missing dependency 'node-html-parser'. Run bun install in this skill directory.`
+
+`--url` is **required**. There is no brand-name search fallback — omitting it
+fails immediately and names the flag.
 
 ## Usage
 
 ```bash
-skills run brand-assets --url https://example.com
-skills run brand-assets --brand "Acme Ledger" --sizes 64,128,256,512,1024
-skills run brand-assets "Acme Ledger" --url https://example.com --include-screenshot
+# the normal case
+skills run brand-assets -- --url https://example.com
+
+# positional URL, custom destination and timeout
+skills run brand-assets -- https://example.com --output ./out/example --timeout 30000
+
+# inspect without downloading any binaries
+skills run brand-assets -- --url https://example.com --no-download
+
+# machine-readable profile on stdout
+skills run brand-assets -- --url https://example.com --json
 ```
 
+Run it directly from the skill directory with `bun run src/index.ts --help`.
+
 ## Options
 
 | Option | Description | Default |
 |--------|-------------|---------|
-| `--brand <text>` | Brand, product, or company name. Positional text also works. | inferred when possible |
-| `--url <url>` | Official website or landing page to inspect. | optional |
-| `--sizes <list>` | Comma-separated PNG export sizes. | `64,128,256,512,1024` |
-| `--include-screenshot` | Include a visual reference screenshot when available. | false |
+| `--url <url>`, `-u` | Website to inspect. A bare hostname is upgraded to `https://`. A positional URL also works. | **required** |
+| `--output <dir>`, `-o` | Output directory. | `./brand-assets` |
+| `--timeout <ms>`, `-t` | Per-request timeout, 500–300000. Applies to the page, each stylesheet, and each asset. | `15000` |
+| `--max-assets <n>` | Cap on downloaded assets, 0–200. Anything beyond is listed as `skipped`. | `20` |
+| `--no-download` | Discover and report only; fetch no binaries. | off |
+| `--json` | Print the full profile as JSON on stdout. | off |
+| `--help` | Show help and exit 0. | |
+| `--version` | Print the version. | |
+
+## What is discovered
+
+| Source | What is taken |
+|--------|---------------|
+| `<link rel="icon">`, `shortcut icon`, `apple-touch-icon`, `mask-icon` | Favicons and touch icons, with their `sizes` and `type` |
+| `<link rel="manifest">` | The web app manifest: `name`, `short_name`, `theme_color`, `background_color`, and every declared icon |
+| `<meta property="og:image">`, `<meta name="twitter:image">` | Social preview images |
+| `<meta name="theme-color">` | Brand color, including per-`media` light/dark variants |
+| `<img>` whose `src`, `alt`, `class`, or `id` matches logo / wordmark / brandmark / brand | Logo images |
+| Inline `<svg>` inside `header`, `nav`, links, or logo-classed elements | Up to three inline marks, saved as standalone SVG files |
+| Inline `<style>`, `style=` attributes, and up to 6 linked stylesheets | CSS custom properties whose name mentions color / brand / accent / primary / bg / fg / surface / theme, plus ordinary `color`, `background`, `fill`, and `stroke` declarations |
+| The same CSS | `font-family` declarations and font-related custom properties |
 
-## Hosted Workflow
+If the page declares no icon at all, `/favicon.ico` is tried as a conventional
+fallback and reported honestly if it 404s.
 
-1. Resolve the most likely official website and brand pages from the supplied brand or URL.
-2. Inspect the selected pages for linked logos, favicon assets, social preview images, brand colors, fonts, typography, and structured metadata.
-3. Prefer official vector logos, then normalize fallback image assets into the requested PNG sizes.
-4. Name every asset with a stable brand slug, file role, variant, and size.
-5. Include source URLs and confidence notes so downstream users know where each asset came from.
+### How colors are ranked
+
+`<meta>` and web app manifest colors first, then hand-authored CSS custom
+properties, then ordinary declarations. Utility-framework plumbing (`--tw-*` and
+friends) is deliberately demoted so real brand tokens surface first. Every entry
+records how many times it was seen and every source it appeared in. The palette
+is capped at 48 distinct colors.
 
 ## Outputs
 
-- `manifest.json`
-- `brand-profile.json`
-- `brand-profile.md`
-- `palette.json`
-- `typography.md`
-- `logo-usage.md`
-- `sources.json`
-- `logos/svg/<brand>-logo.svg`
-- `logos/png/<brand>-logo-64.png`
-- `logos/png/<brand>-logo-128.png`
-- `logos/png/<brand>-logo-256.png`
-- `logos/png/<brand>-logo-512.png`
-- `logos/png/<brand>-logo-1024.png`
-
-After submitting a hosted run, poll with `skills runs status <run-id>` and download outputs with `skills exports download <run-id>`.
+| File | Contents |
+|------|----------|
+| `assets/` | Every downloaded file, named `<role>-<original-name>.<ext>` — e.g. `icon-favicon-32x32.png`, `logo-img-wordmark.svg`, `inline-logo-01.svg`. |
+| `brand-profile.json` | Everything discovered: metadata, assets with status, palette, typography, stylesheet results, counts. |
+| `brand-profile.md` | The same profile as a readable summary with asset, palette, and typography tables. |
+| `palette.json` | Every color: hex, how it was declared, which properties carried it, its kind, occurrence count, and every source. |
+| `typography.md` | Each font stack with its full fallback chain, source, and notes on reproducing the type without the font binaries. |
+| `sources.json` | The provenance map: downloaded file → source URL, status, content type, byte size, plus per-stylesheet fetch results. |
+| `manifest.json` | Every file written, with byte sizes and the run inputs. |
+
+Webfont binaries are never downloaded — licenses almost never permit
+redistributing them. `typography.md` gives you the stacks instead.
+
+## Failure behaviour
+
+Exit code 1 with a single clear line on stderr for each of:
+
+- `--url` missing — names the flag and shows an example.
+- A value that is not an http(s) URL.
+- DNS or connection failure.
+- A timeout — the message tells you to raise `--timeout`.
+- A non-200 response — the status code is included.
+- A response whose content type is not HTML — the actual content type is quoted.
+
+Failures *below* the page level are not fatal: a 404 favicon, an unreachable
+stylesheet, or a malformed manifest is recorded with its status in
+`sources.json` and `brand-profile.md`, and the run still completes.
+
+## Provenance and reuse
+
+These are someone else's trademarks. `sources.json` exists so you can prove where
+each file came from. Confirm you have the right to use an asset before you ship
+it, and check the site's own brand or press page for official guidelines and
+higher-resolution originals.

package/skills/brand-assets/package.json

@@ -2,10 +2,34 @@
   "name": "brand-assets",
   "version": "0.1.0",
   "private": true,
-  "description": "Hosted premium brand asset discovery package",
+  "description": "Extract brand assets, colors, and typography from a website you name, with full source provenance",
   "type": "module",
-  "skills": {
-    "runtime": "hosted",
-    "source": "remote"
+  "main": "src/index.ts",
+  "bin": {
+    "brand-assets": "src/index.ts"
+  },
+  "files": [
+    "src",
+    "SKILL.md",
+    "tsconfig.json"
+  ],
+  "keywords": [
+    "brand",
+    "logo",
+    "favicon",
+    "palette",
+    "scraper",
+    "skill"
+  ],
+  "scripts": {
+    "start": "bun run src/index.ts",
+    "typecheck": "tsc --noEmit"
+  },
+  "dependencies": {
+    "node-html-parser": "^6.1.13"
+  },
+  "devDependencies": {
+    "@types/bun": "latest",
+    "typescript": "^5.7.0"
   }
 }

package/skills/brand-assets/src/cli.ts

@@ -0,0 +1,155 @@
+import { VERSION } from "./constants.js";
+import type { CliOptions, HtmlNode } from "./types.js";
+
+export async function loadHtmlParser(): Promise<(html: string) => HtmlNode> {
+  try {
+    const mod = await import("node-html-parser");
+    return mod.parse as unknown as (html: string) => HtmlNode;
+  } catch {
+    throw new Error("Missing dependency 'node-html-parser'. Run bun install in this skill directory.");
+  }
+}
+
+/* ------------------------------------------------------------------ */
+/* cli                                                                 */
+/* ------------------------------------------------------------------ */
+
+function printHelp(): void {
+  console.log(`brand-assets v${VERSION}
+
+Extract brand assets from a website you name. Fetches the page over plain HTTP,
+parses it for icons, social preview images, theme colors, CSS color custom
+properties, and font stacks, then downloads what it finds.
+
+No API keys and no third-party services: the only host contacted is the one in
+the URL you pass (plus whatever that page links its own assets from).
+
+USAGE:
+  brand-assets --url <url> [options]
+  brand-assets <url> [options]
+
+OPTIONS:
+  -u, --url <url>           Website to inspect. Required.
+  -o, --output <dir>        Output directory                     [./brand-assets]
+  -t, --timeout <ms>        Per-request timeout in milliseconds  [15000]
+      --max-assets <n>      Maximum assets to download           [20]
+      --no-download         Discover and report only, fetch no binaries
+      --json                Print the brand profile as JSON on stdout
+      --help                Show this help message
+      --version             Show the current version
+
+WHAT IS DISCOVERED:
+  <link rel="icon" | "shortcut icon" | "apple-touch-icon" | "mask-icon">
+  <link rel="manifest">      (parsed for name, theme_color, background_color, icons)
+  <meta property="og:image">, <meta name="twitter:image">
+  <meta name="theme-color">  (including per-color-scheme media variants)
+  <img> and inline <svg> whose src, alt, class, or id looks like a logo
+  CSS custom properties whose name contains color / brand / accent / bg / fg
+  font-family declarations, in inline <style> and in linked stylesheets
+
+OUTPUTS:
+  assets/                   Downloaded icons, logos, and social images
+  brand-profile.json        Everything discovered, machine readable
+  brand-profile.md          The same profile as a readable summary
+  palette.json              Every color with its hex and where it was found
+  typography.md             Font stacks with their source
+  sources.json              asset file -> source URL map
+  manifest.json             Every file written, with byte sizes
+
+EXIT CODES:
+  0  page fetched and parsed
+  1  missing --url, network failure, non-200 response, or non-HTML content type
+
+EXAMPLES:
+  brand-assets --url https://example.com
+  brand-assets https://example.com --output ./out/example --timeout 30000
+  brand-assets --url https://example.com --no-download --json
+`);
+}
+
+export function parseArgs(argv: string[]): CliOptions {
+  const options: CliOptions = {
+    output: "./brand-assets",
+    timeout: 15000,
+    maxAssets: 20,
+    download: true,
+    json: false,
+  };
+
+  for (let i = 0; i < argv.length; i += 1) {
+    const arg = argv[i];
+    switch (arg) {
+      case "--help":
+      case "-h":
+        printHelp();
+        process.exit(0);
+      case "--version":
+      case "-v":
+        console.log(VERSION);
+        process.exit(0);
+      case "--url":
+      case "-u":
+        options.url = argv[++i];
+        break;
+      case "--output":
+      case "-o":
+        options.output = argv[++i] ?? options.output;
+        break;
+      case "--timeout":
+      case "-t": {
+        const value = Number.parseInt(argv[++i] ?? "", 10);
+        if (!Number.isFinite(value) || value < 500 || value > 300000) {
+          throw new Error(`Invalid --timeout value: ${argv[i]} (expected 500-300000 ms)`);
+        }
+        options.timeout = value;
+        break;
+      }
+      case "--max-assets": {
+        const value = Number.parseInt(argv[++i] ?? "", 10);
+        if (!Number.isFinite(value) || value < 0 || value > 200) {
+          throw new Error(`Invalid --max-assets value: ${argv[i]} (expected 0-200)`);
+        }
+        options.maxAssets = value;
+        break;
+      }
+      case "--no-download":
+        options.download = false;
+        break;
+      case "--json":
+        options.json = true;
+        break;
+      default:
+        if (arg.startsWith("-")) {
+          throw new Error(`Unknown option: ${arg}`);
+        }
+        if (!options.url) {
+          options.url = arg;
+          break;
+        }
+        throw new Error(`Unexpected argument: ${arg}`);
+    }
+  }
+
+  if (!options.url || options.url.trim() === "") {
+    throw new Error(
+      "Missing required --url <url> argument. This skill inspects a specific website; " +
+        "there is no brand-name search fallback. Example: brand-assets --url https://example.com",
+    );
+  }
+
+  return options;
+}
+
+export function normalizeUrl(input: string): URL {
+  const candidate = /^https?:\/\//i.test(input) ? input : `https://${input}`;
+  try {
+    const url = new URL(candidate);
+    if (url.protocol !== "http:" && url.protocol !== "https:") {
+      throw new Error("unsupported protocol");
+    }
+    return url;
+  } catch {
+    throw new Error(`Invalid --url value: ${input} (expected an http(s) URL such as https://example.com)`);
+  }
+}
+

package/skills/brand-assets/src/constants.ts

@@ -0,0 +1,8 @@
+export const VERSION = "0.1.0";
+export const USER_AGENT =
+  "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0 Safari/537.36";
+export const MAX_ASSET_BYTES = 8 * 1024 * 1024;
+export const MAX_STYLESHEETS = 6;
+export const MAX_STYLESHEET_BYTES = 2 * 1024 * 1024;
+export const MAX_PALETTE_ENTRIES = 48;
+