download_file
Native HTTP/HTTPS downloads with redirects, headers, progress and integrity checks.
| Parameter | Notes |
|---|---|
| url | HTTP/HTTPS address |
| path | target path (workspace-relative or global path) |
| destination | workspace / global |
| headers | e.g. Authorization: Bearer xxx |
| sha256 | optional integrity check, fails on mismatch |
| timeout_seconds | download timeout |
TIP
Prefer download_file over curl/wget - progress, redirects and verification built in.
web_search
Web search returning candidate titles/URLs/snippets. Snippets are leads, not evidence.
read_web_page
Open a page and read the body.
WARNING
- prefer official docs and authoritative sources;
- never bypass a user-blocked page;
- switch sources or state the limitation for login walls / dynamic rendering / too-short pages.
mark_web_sources
Record the sources actually used and include links in the answer. Call it only when web content was used.
Website blocklist
The blocklist in Settings stops AI from opening domains:
- exact match:
x.comblocks onlyx.com;www.x.commust be added separately; - wildcard:
*.x.comblocks all subdomains; - full block: add both
x.comand*.x.com.
