An Agent Skill that converts websites and PDFs into clean, LLM-friendly markdown using Jina Reader. Saves the markdown to local files by default, or loads it directly into the agent's session via curl.
Copy the jina-reader/ directory into your agent's skills folder:
| Tool | Skills location |
|---|---|
| Claude Code | ~/.claude/skills/ |
| Codex | ~/.agents/skills/ |
| Gemini CLI | ~/.gemini/skills/ |
| VS Code / Copilot | .github/skills/ in your repo |
Example for Claude Code:
git clone <repo-url> /tmp/jina-reader-skill
cp -r /tmp/jina-reader-skill/jina-reader ~/.claude/skills/Optional but recommended — set a Jina API key for higher rate limits (500 RPM vs the 20 RPM free tier):
export JINA_API_KEY="your-key-here"Get a key for free at jina.ai/reader.
Once installed, prompt your agent naturally:
- "Download this page as markdown — https://en.wikipedia.org/wiki/Solaris_(1972_film)"
- "Download all URLs in
movies.txtas markdown" - "Load this page into context as markdown, article only"
The skill picks the right path automatically: the bundled Python script for save-to-disk asks, curl for load-into-context.
Four presets cover most use cases:
default— clean markdown, strips nav/header/footer/aside and imagesfull-page— full page as markdown, nothing strippedtext-only— main content only, no images or linkswait-for-content— waits up to 30s for JS-rendered content
For finer control, the script exposes selector and timeout overrides — see jina-reader/SKILL.md.
--insecureis enabled by default to avoid SSL friction on Python setups without certificates installed. Disable with--no-insecureif your environment is locked down.
MIT