v0.1.0 - initial release

web-fetch v0.1.0

LLM-neutral skill that fetches a URL to a temp file so agents can pipe the path through rg / jq / awk instead of dumping the whole page into context.

Highlights

• Single-file Python 3 stdlib (no pip install)
• curl preferred with urllib fallback
• HTML → Markdown via html.parser (no Turndown / pandoc dependency)
• Cloudflare-aware retry (browser UA → honest UA on 403/503 + body markers)
• Per-attempt trace in result.json for auditability
• Cross-OS: macOS, Linux, Windows; tested on Python 3.9 → 3.13

Behavior

Same shape as pi-webfetch and opencode/webfetch:

• One GET, one Cloudflare retry, save raw + rendered + envelope to $TMPDIR/web-fetch-<runid>/, print the path on stdout.
• 5 MB response cap, 120 s timeout cap, public HTTP/HTTPS only.
• No providers, no fallback chain, no JS rendering, no auth - by design.

Install

git clone https://github.com/code-yeongyu/web-fetch ~/.agents/skills/web-fetch

Or symlink for active development:

ln -s /path/to/your/clone ~/.agents/skills/web-fetch

Tested matrix

• OS: macOS (latest), Ubuntu 22.04, Ubuntu (latest), Windows (latest)
• Python: 3.9, 3.10, 3.11, 3.12, 3.13
• Total CI cells: 20 smoke + 2 syntax-floor + 1 lint = 23 (all green)