[Improve] Show Auto more of what the agent read

[daniel-lxs]

Why

Auto was shown the newest 8 tool results and the first 1,500 characters of each. When the agent looked a person or an item up in a longer listing, or made a few more calls before using the id it found, the lookup was cut off or dropped. The model then could not see what the id referred to and asked for approval, for example on "assign this to me" where the tool takes a user id.

What

• Auto is now shown the newest 20 results, up to 6,000 characters of each and 16,000 across them, counting each result's arguments as well as its output.
• With whole listings in view, the owner wording now says that a record whose name or address only resembles the owner's belongs to somebody else.

How to test

1. Turn Auto on for a session with a tracker that takes assignees as user ids, in a workspace whose user list is long.
2. Ask "Assign ISSUE-1 to me."
3. The agent looks up your user id and assigns the issue without an approval card.

A unit test covers the new limits.

[roomote-community]

No code issues found. See task

• packages/cloud-agents/src/server/integration-tool-auto-identifiers.ts:12 Tool-result arguments are excluded from the new shared context budget, allowing a 20-result prompt to substantially exceed 16,000 characters.
• packages/cloud-agents/src/server/integration-tool-auto-identifiers.ts:50 Tool names are excluded from the shared context budget, so a 20-result prompt can still exceed 16,000 characters.

_{Reviewed 55dc1e5}