fix(agent): tell the model that navigating a page is not reading it

[MummIndia]

Summary

browser_navigate returns the page title and a snapshot reference. It never returns
the page text. Asked to open a URL and summarise it, the agent called
browser_navigate, saw exit_code=0, answered "the page loaded successfully" and
stopped — having read nothing. The turn looks successful and the tool log agrees, so
nothing flags it.

browser_snapshot has the same trap one level down: its optional filename argument
is documented as "Save snapshot to markdown file instead of returning it in the
response", and the model volunteers one. It then receives a path it cannot open —
175 characters of reference where omitting the argument returns 1634 of content.

This adds a note covering both. Navigation is never the last step, follow it with
browser_snapshot or browser_find; call snapshot with no arguments; and prefer
web_fetch when the task is only to read a page, since it does the whole thing in
one call rather than three round trips — which matters on local hardware where each
round trip is a full generation.

The note is injected only when browser tools are in that turn's selection, and is
suppressed alongside the other local-context blocks, so it costs nothing on unrelated
requests.

Target branch

• This PR targets dev, not main.

Linked Issue

Fixes #5764

Type of Change

• Bug fix (non-breaking — fixes a confirmed issue)

Checklist

• I searched open issues and open PRs — this is not a duplicate.
• This PR targets dev
• My changes are limited to the scope described above — no unrelated refactors or whitespace changes mixed in.
• I actually ran the app (docker compose up) and verified the change works end-to-end.

How to Test

1. Run with the built-in browser MCP server connected, in agent mode.

2. Send: Open https://example.com and summarise what this page is about.

3. Before the change — the agent calls browser_navigate, then ends the turn with
   a message along the lines of "the page loaded successfully", without answering the
   question. The round summary reads 0 native calls, 0 tool blocks.

4. After the change — the agent follows browser_navigate with browser_snapshot
   and answers from the page content. Confirmed in the logs:

   Tool executed: mcp__builtin_browser__browser_navigate -> exit_code=0
   Tool executed: mcp__builtin_browser__browser_snapshot -> exit_code=0
   

5. To check the conditional injection, build the system prompt with and without a
   browser tool in relevant_tools. The READING A WEB PAGE block is present in the
   first case and absent in the second.

Visual / UI changes

None. This PR only touches src/agent_loop.py — prompt text assembled server-side.
No HTML, CSS, SVG or static/js/ module is modified.