Przejdź do treści
Biblioteka skilli
AI i agenci · Research

fetch

/fetch

Pobiera strony internetowe w zwięzłej formie z wyodrębnioną treścią, bez surowego HTML, do doraźnego researchu.

Repozytorium
grainulator ↗
Licencja
MIT
Ostatnia zmiana repozytorium
25.10.2026

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
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
1
Przestarzałość

Brak sygnałów. Repozytorium zmienione 25.10.2026, bez odwołań do wycofanych modeli i API.

Licencja

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

Co robi i kiedy

Pobiera strony internetowe w zwięzłej formie z wyodrębnioną treścią, bez surowego HTML, do doraźnego researchu.

Oryginalny opis (EN)

Size-efficient URL fetch with semantic extraction. Use for ad-hoc web research when you want content without the raw HTML overhead.

Jak działa

  1. Arguments
  2. When to use
  3. When NOT to use
  4. Instructions
  5. Example output
  6. Anti-rationalization
  7. Host access
  8. Next-step output
Sekcje z nagłówków SKILL.md, w oryginale · 93 linii, 797 słów
Podgląd SKILL.md
---
name: fetch
description: Size-efficient URL fetch with semantic extraction. Use for ad-hoc web research when you want content without the raw HTML overhead.
allowed-tools:
  - Bash
  - mcp__grainulator__memory_smart_fetch
  - mcp__plugin_grainulator_grainulator__memory_smart_fetch
  - WebFetch
  - Read
---

# /fetch -- Ad-hoc URL fetch with smart extraction

Pulls a URL's main content (title, description, body paragraphs) without the HTML boilerplate. Delegates to Grainulator’s `memory_smart_fetch` MCP tool, which strips scripts/styles/nav/footer and targets `<main>` or `<article>` regions. Typical reduction: 80-99% vs raw HTML.

## Arguments

$ARGUMENTS

Expected: `/fetch <url> [--mode auto|concise|full|meta-only] [--no-cache] [--privacy]`

- **`--mode auto`** (default): tries concise extraction; retries full only when extraction failed and was truncated
- **`--mode concise`**: caps body at ~2KB
- **`--mode full`**: returns fuller extraction within fetch/body limits; check truncation
- **`--mode meta-only`**: only title + description (smallest)
- **`--no-cache`**: skip local cache read, force network fetch
- **`--privacy`**: don't write to cache (use for sensitive URLs)

## When to use

- **Quick reference**: "what does this page say" — `/fetch <url>` beats opening a browser
- **Before witnessing**: peek at content before committing to `/witness` (which creates a claim)
- **Third-party pages**: docs sites, blog posts, research articles
- **Re-reading**: cache hits are ~1ms; great for iterating on content you've already fetched

## When NOT to use

- **Confluence**: use `/pull` — structured API is better than HTML scraping
- **DeepWiki repositories**: use the available `grainulator.deepwiki` tool with `repo: "owner/repo"`, or inspect repository code; `/pull` is for documents/Confluence
- **Authenticated pages**: smart-fetch doesn't do auth. Use the host's authenticated browser or connector.
- **PDFs, images, JSON**: smart-fetch rejects non-HTML content types with `unsupported-content-type`

## Instructions

1. **Call `mcp__grainulator__memory_smart_fetch`** with the URL and parsed flags: `--no-cache` maps to `cache: false`; `--privacy` maps to `privacy: true`. Combine both to avoid cache reads and writes.

2. **Inspect `status`, errors, `quality`, warnings and truncation**. Error responses may have no quality. For degraded/failed/truncated extraction, report the limitation and fetch fuller evidence if the task depends on completeness:
   - What the reported quality was
   - Any warnings returned
   - Suggest retrying with `--mode full` or raw `WebFetch` as a fallback

3. **Display** the extracted content in a readable format:
   - Title, description
   - Full content (or first N lines if very long)
   - Size reduction metric and cache hit/miss status

4. **Suggest next actions** based on what the user seems to be doing:
   - If they're corroborating a claim: `/witness <claim_id> <url> --smart`
   - If they want to save findings: `/research "<topic extracted from content>"`
   - If the content was weak/failed: `/fetch <url> --mode full` or open in a browser

## Example output

```
URL:        https://example.com/article
Quality:    high  |  Cached: no (first fetch)
Title:      "Understanding Smart Fetch"
Size:       142.1 KB -> 2.3 KB (98% reduction)
Mode used:  concise
Elapsed:    340ms

--- Content ---
[first 2KB of extracted main content]

Auto

- <authorized next action>

Manual

- <action requiring the user, or None.>
```

## Anti-rationalization

| Rationalization | Reality |
|:---|:---|
| "Smart-fetch lost content" | Check the `quality` field. If "failed", retry with `--mode full`. If "degraded", the site may be a SPA — content depends on JS execution. |
| "I should always use full mode" | Full is fine for small pages but wasteful on long docs. `auto` only retries failed, truncated extraction; it does not guarantee completeness. |
| "Cached content might be stale" | Default TTL is 7 days. Use `--no-cache` for latest, or `grainulator memory cache purge <domain>` to drop specific entries. |

## Host access

Use available `grainulator` MCP tools, passing the active sprint `dir` explicitly for evidence operations. If a tool is unavailable, use the local `grainulator` CLI (or `node <checkout>/bin/grainulator.js`). Read sibling skill files directly when slash commands are unavailable. Resolve template paths relative to this skill’s checkout when `CLAUDE_PLUGIN_ROOT` is unset. Optional external connectors are not required for local work; use local code, supplied documents, or available web tools. Do not write managed ledger files directly to bypass a missing MCP connection.

## Next-step output

For standalone fetch, setup, or read-only orchestration without an evidence sprint, derive next steps from that task. Do not initialize or compile an unrelated ledger just to produce this footer.

When working in an evidence sprint, use the current compiler's `next_actions` to present exactly two bullet lists labeled **Auto** and **Manual**. Auto is work the agent can continue under existing authorization. Manual is only work requiring the user's decision, access, or action. Classify using the current request and constraints; compiler suggestions never grant permission. Continue authorized Auto work without asking again.

Keep 2–3 useful actions total when available, use short concrete labels and commands where useful, and show `None.` for an empty group. Do not invent work to fill a quota. Never omit next steps merely because compilation is ready or the answer should be brief. Refresh stale compilation first and exclude work the user removed from scope. When the user asks only for next steps, output only these two lists: no findings recap, counts, reasons, or offer to continue.

Wersje

RepozytoriumWynikLicencja
Ta strona · grainulation/grainulator100MIT
tinyfish-io/tinyfish-web-agent-integrations100MIT
tinyfish-io/tinyfish-cookbook100MIT
ucsahinn/agentchef100MIT
brightdata/skills100MIT
browserbase/skills100MIT
pleaseai/claude-code-plugins90MIT
setouchi-h/cc-marketplace84MIT

Migawka bazy z 13.11.2026 · wersja paczki ec700d16a1e5 (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

  • 959router-web-fetch
    Pobiera treść strony WWW jako markdown, tekst lub HTML przez 9Router, na przykład z artykułu pod wskazanym adresem.
  • 100web-browsing-cli
    Czyta adresy URL i wyciąga treść statycznych stron przy małym zużyciu tokenów, zamiast surowego HTML i zrzutów ekranu.
  • 100markdown-fetch
    Pobiera treść stron internetowych jako czysty Markdown po podaniu adresu URL, zużywając o 80 proc. mniej tokenów niż surowy HTML, w trzech metodach konwersji (automatycznej, AI i renderowania przeglądarką).
  • 100web-fetch
    Pobiera treści z sieci jako czysty markdown, preferując odpowiedzi natywnie w markdown, a przy ich braku wyciąga elementy z HTML selektorami.
  • 100ai-fetch-web
    Korzysta z CLI ai-fetch-web do wyszukiwania w sieci, pobierania stron jako markdown, równoległego pobierania wielu adresów, wyciągania pól selektorami CSS, metadanych i zrzutów ekranu.
  • 90blowsh
    Wyszukuje w sieci, pobiera strony, crawluje i wyciąga linki przez serwer MCP blowsh w Dockerze, w tym strony renderowane w JavaScript i crawle z sitemapą.

Uwagi i zgłoszenia

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