web-search-tor

web-search-tor

Everything in the web-search skill, over Tor. Read that skill first for the commands; this one covers the layer, what changes when it is on, and what does not work without it.

Turn it on

websearch tor up          # start Tor, verify the exit, turn the layer on
websearch tor status      # is it listening, and is the traffic really Tor
websearch tor down        # stop it and turn the layer off

up is a one-call bring-up: it uses a Tor that is already listening, else tor on PATH, else it downloads the official Tor Expert Bundle into the state directory and checks it against the sha256 published beside it. Then it starts Tor detached, waits for the bootstrap, and asks check.torproject.org whether the traffic really leaves through Tor. Give it a timeout_s of 300 on the first run.

The layer is written to your settings file, so it stays on for later commands until websearch tor down. websearch tor status is the check to run before trusting a session: a port that answers is not the same thing as a port that is Tor, and the is_tor field is the one that says which.

What changes while it is on

• Every request the tool makes goes through Tor: search, fetch, arxiv, github.
• .onion addresses work. With the layer off they are refused before anything resolves, because looking up an onion name locally leaks it and fails anyway.
• --onion searches the onion indexes instead of the clearnet engines.
• An egress proxy you already had configured is not dropped. It becomes Tor's upstream, so the path is you, then the proxy, then Tor. Your ISP sees the proxy rather than a Tor entry guard.

Search the onion network

websearch web-search "<query>" --onion [--max-results 8]
websearch search "<query>" --onion [--max-results 20] [--freshness week]

--onion replaces the engine set rather than extending it: no clearnet engine indexes onion services, and no onion index crawls the clearnet, so a mixed fanout would be two half-empty result lists fused over disjoint corpora. Results are ordinary results with .onion URLs and handles, so web-fetch and web-open take them unchanged.