[Feature]: Quote a selection from an assistant reply as context for the next prompt

[MarkusReichart]

Before submitting

• I searched existing issues and did not find the same request.

(Related: #798 asked for essentially this but was closed by the stale bot without discussion; #1170 and #1311 cover neighbouring use cases. See below.)

Problem or unmet use case

Selecting text in a note and having it automatically attached as <editor_selection> is one of my favourite Claudian features. It makes it effortless to point the agent at an exact passage.

Inside the chat panel, though, this doesn't work. I often want to follow up on one specific sentence in an earlier assistant reply. Today I have to copy it, paste it into the input, format it as a quote and then type my question. It's friction-heavy, it breaks the conversational flow, and in long replies the agent sometimes guesses wrong about which part I mean when I paraphrase instead of quoting.

Desired outcome

Text selected in an assistant reply behaves like text selected in a note: it is attached to my next prompt as context, so I can ask a targeted follow-up about exactly that passage. The attached selection is visible in the composer (like the existing selection indicator) and can be removed before sending.

Affected provider

Not provider-specific

Proposed solution

Reuse the existing selection-context mechanism:

• Selecting text inside an assistant message captures it as a <chat_selection> context block, analogous to <editor_selection> and <browser_selection>.
• Optionally, show a small floating "Quote" / "Add to follow-up" button near the selection instead of attaching it automatically, so normal copy/paste selections aren't attached by accident.
• Nice to have, not required: identifying which reply the selection came from, and supporting selections in user messages too.

Minimal v1: select text in an assistant reply → it is shown as context in the composer → it is sent with the next prompt.

Alternatives considered

• Manual copy, paste and blockquote: works, but takes four steps instead of one and interrupts the flow.
• Paraphrasing ("regarding your second point…"): ambiguous in long replies.

Additional context

• Feature: Quote selection from chat history (Perplexity-style) #798 requested the same thing (Perplexity-style quoting); it got no maintainer response before the stale bot closed it, so I think there is still demand.
• [Feature]: Explain a term selected in an assistant reply without leaving the chat #1170 (ask about a selected term in a popover) and [Feature]: Response Annotation Queue #1311 (annotation queue) are related but broader: they add separate UI (popover or annotation list). This request only reuses the existing selection-context flow, the smallest building block those features could share.
• Claudian 2.3.3, Obsidian desktop, Linux.