feat(browser): adblock on|off|status — the per-site off switch for pinned uBlock Origin Lite (DIVE-4516)

[5dive-bot]

What

5dive browser adblock on|off|status <site> — the per-site off switch for the
pinned uBlock Origin Lite that DIVE-4516's stack half installs into agent Chrome
profiles by managed policy. The plugin half of that row.

5dive browser adblock status                 # is it on, and which sites is it off for
sudo 5dive browser adblock off example.com   # this site breaks under filtering — stop filtering it
sudo 5dive browser adblock on  example.com   # filter it again

The premise was measured before it was built on

uBOL's own per-site switch lives in its popup and is not scriptable from a seat.
The one that is, is Chrome policy's ExtensionSettings.<id>.runtime_blocked_hosts
— and that key is documented against content-script injection and extension API
access, while uBOL Lite filters through declarativeNetRequest, a static
ruleset evaluated in the network stack. Different mechanisms; the row said measure
it rather than assume it.

Measured on exact-swallow at Chrome 153 (receipt on the row, 2026-09-14): with the
key set for the host, the DNR block disappeared (0 → 3000 divs, 149 B → 65002 B)
and the content script stopped injecting; removing it restored both. It is a
real off switch, and it is total for that host.

Four decisions worth reviewing

• Root, like setup. Chrome policy on Linux is machine-level only — there is no
  per-user policy path — so adblock joins setup as the second verb a root caller
  is not dropped out of at the bottom of bin/browser. T2c8 asserts no third
  verb quietly joins them.
• The off list is the source of truth, not the policy file.
  /var/lib/5dive/browser/ubol/adblock-off. The nightly root converge re-renders
  the whole policy file; a site turned off at 14:00 that lived only in that file
  would be silently re-filtered at 03:00, with nobody to notice.
• Both host patterns. *://*.example.com does not match example.com, so a
  wildcard-only off switch reports success and leaves the apex — the host the seat
  actually typed — still filtered. Mutating this to wildcard-only reds 3 arms.
• Live pickup is not promised. Only fresh launches were measured, so the verb
  says shot/read take it on their next render and a browser already running
  under serve may need serve <site> --stop.

A box with no policy file still records the setting and says so plainly, rather than
reporting an off switch it did not apply.

Checked

tests/browser_plugin_unit.sh — 531 arms, 531 pass (513 before, 18 new).
Mutation-checked on this tree: a wildcard-only off switch reds 3; leaving an empty
array instead of deleting the key reds 1. The non-root refusal is covered by a
paired positive/negative (T91 refuses at exit 77, T93 succeeds under a root
caller), not by a mutant — the mutant patch for it did not apply cleanly and is not
claimed here.

Plugin version 1.6.0. README and the connect-site skill both say when to reach for
it. Not raising _BS_PLUGIN_MIN in 5dive-api: that is the fleet rollout knob,
and it should turn when the CRX this verb switches off is actually on the boxes.

🤖 Generated with Claude Code