web-fetch

web-fetch is a skill for Claude Code from aksika/abtars. It costs 17 tokens per session (736 once invoked), scanned A, original, Apache-2.0.

A web-content fetching skill that retrieves pages using progressively more capable methods, from raw HTML downloads to an interactive browser.

In plain words
What is it for?
Fetching readable text from websites, loading JavaScript-rendered pages, and opening pages that need browser interaction.
Why use it?
It helps when a page is protected, cluttered, or built with JavaScript and a simple download cannot provide usable content.

Skill for Claude Code

Written for Claude Code: user-invocable in frontmatter.

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/aksika/abtars/web-fetch
Any agent
npx skills add aksika/abtars --skill web-fetch
Clone the repo
git clone --depth 1 https://github.com/aksika/abtars

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for web-fetch

README.md
[![agentmods](https://agentmods.dev/badge/skills/aksika/abtars/web-fetch.svg)](https://agentmods.dev/skills/aksika/abtars/web-fetch)
Your own site
<a href="https://agentmods.dev/skills/aksika/abtars/web-fetch"><img src="https://agentmods.dev/badge/skills/aksika/abtars/web-fetch.svg" alt="Measured on agentmods" height="20"></a>
Per session 17 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 736 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00017 $0.00736
Opus 5 $0.00009 $0.00368
Sonnet 5 $0.00003 $0.00147
Haiku 4.5 $0.00002 $0.00074

Measured 6d ago against content hash 100ced183286, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

web-fetch scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

description: Fetch web content — 4-level escalation chain from curl to full browser
templates/skills/web-fetch/SKILL.md · 99 lines

How it starts

The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.

web-fetch

Four levels of web content retrieval. Use the lightest level that works. Escalate up only when the current level fails.

Level 1 — curl (fastest, no rendering)

curl -sL "URL" --max-time 10 | head -200

Use when: open site, simple HTML, just need raw text.

Level 2 — Jina Reader (CF bypass, clean markdown)

curl -sL "https://r.jina.ai/URL" \
  -H "User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36" \
  --max-time 15

Use when:

  • Direct curl returns CF challenge page or HTML soup
  • Want clean markdown output instead of raw HTML
  • Site has basic Cloudflare protection

Limitations: no JS execution, rate-limited (don't loop >5 URLs rapidly), some aggressive CF still blocks it.

Level 3 — lightpanda (JS rendering, fast)

lightpanda fetch --dump markdown --strip-mode full --wait-ms 5000 "URL"

Use when:

  • Levels 1-2 returned empty/broken content
  • Page requires JavaScript to render (SPA, React, dynamic content)
  • Still a single page, no interaction needed

10x faster than Chrome. Handles most JS-rendered pages. No login/interaction.

Level 4 — CloakBrowser (full browser, interaction)

Escalate to a managed Browsie session. Create a B-type kanban card with a detailed goal and report the card ID. Do NOT invoke a browser CLI from Main.

Goal checklist

Include in the goal:

  • Objective and expected outcome
  • Starting URL(s) and any prior fetch results
  • Required interactions (login, forms, clicks, navigation)
  • Expected artifacts (screenshots, extracted text, page info)
  • Success criteria and stopping condition

Direct API

kanban_manage action=create type=B title="<short title>" goal="<detailed goal>"

ACP

abtars kanban create --type B --title "short title" --goal "$(cat <<'GOAL'
detailed goal text
GOAL
)"

Or pass a goal file:

abtars kanban create --type B --title "short title" --goal-file /tmp/goal.txt

Acknowledgement

When the card is queued, say: "Browsie task #<id> queued — result will appear here when complete." Do NOT proceed with inline browser commands.

Read the full file on GitHub · 99 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 6d ago First seen · 99 lines · 17 tokens per session scan A 100ced183286

Subscribe to this mod's changes

web-fetch is a skill published in the GitHub repository aksika/abtars (9 stars, last pushed yesterday), licensed Apache-2.0. It adds 17 tokens to every session and 736 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.