Creating another tab for every request scatters accounts, conversations and unsent drafts. I preferred an appropriate existing tab, while defining reuse and reload conditions separately. Existing cannot mean permission to overwrite any manuscript found there.
Distinguish adoption from another navigation
Reloading an existing tab at the same destination can erase page-local input. Reuse returns both a selected page and whether navigation is needed. Equal destinations can continue in place without a redundant reload.
A genuinely different search can navigate the existing surface to the new destination. Reuse does not mean no navigation ever. Preserving state and obtaining a new result require different decisions. Comparing only a reduced display address must not erase the actual query difference either.
Preserve account and resource identity in route matching
Route reuse can accommodate query changes on a search service while retaining parameters identifying accounts and resources. Cafe identifiers, blog-owner paths and different Gemini conversation paths are not casually merged.
Tracking metadata differs from work identity. Removing every parameter makes matching easy but can redirect work to another resource. The classification does not know every future parameter of every service. Another integration must identify which values distinguish account, document and conversation.
Separate program-bound tabs from general candidates
A saved program can locate a tab through an automation key and alias within its project and profile. Explicit multi-tab aliases remain distinct from main, and requests needing another surface retain an exception. Reuse is not an optimization merging every tab into one.
General candidates use default or current selection, exact destinations and unique recency. Some multiple-candidate situations intentionally yield no reuse. The implementation is more specific than the first equal URL. Understanding the applied rule helps avoid handing another document to the wrong task. The actual adoption path also refreshes driver attachment and task ownership. Selecting an ID from metadata and connecting execution control are different stages.
Adoption does not complete manuscript verification
A document alias can adopt an existing editor in place, but adoption does not establish permission to overwrite its contents. Full-manuscript and input checks remain in the workflow. Candidate selection and editor-write policy are separate contracts.
Even a program-bound tab can require navigation when its requested destination differs. The reuse helper alone cannot guarantee preservation of every draft. Execution procedures must protect unsaved text, media and unrelated manuscripts too. Selection convenience and preservation evidence need joint management.
A prohibition is not a request for another tab
Do not open a new tab contains the words new tab without requesting creation. Tests distinguish Korean and English prohibitions from explicit new-surface requests, including a later clause requesting a separate image tab.
Reuse tests cover account and project separation, ambiguous candidates, editor adoption without navigation and navigation to another query. I read the current code, tests and historical usage evidence rather than creating tabs repeatedly in a live account. These rules do not establish perfect interpretation of every positive and negative expression.
The reason for continuity matters more than tab count
Fewer tabs do not automatically improve work. Reuse is valuable when it continues the same task, account and conversation while preserving unrelated manuscripts. Avoiding a new window must not merge work that needs independence.
I prefer existing surfaces with explicit selection boundaries and navigation decisions. The user gains continuity, and the maker can explain why that tab was chosen. That connection keeps agent convenience from becoming a cause of lost work.