Design access_scope declaration strategy for extensions

[Esity]

Context

With the identity scope enforcement work (PR #24, legion-apollo #35, lex-knowledge #15, lex-microsoft_teams #16), we now have:

• Identity always derived from Legion::Identity::Process on writes
• requesting_principal_id auto-injected on reads for filtering
• access_scope: 'private' hardcoded in lex-microsoft_teams

Problem

There's no standard way for extensions to declare what access_scope their data should have. Currently each extension must manually pass access_scope: 'private' on every ingest call. This doesn't scale as more extensions write user-specific data.

Design questions

1. Extension-level default — Should extensions declare default_access_scope :private in their class definition, with all writes from that extension defaulting to private unless overridden?

2. Per-write override — Some extensions have mixed content (e.g., Teams shared channels = team scope, 1:1 chats = private). Do we need per-write granularity on top of the extension default?

3. Both — Extension declares a default, individual writes can override (e.g., :private default for Teams, :team override for shared channel content).

4. Source-based registry — A mapping of source_channel or source_agent to default scope, maintained centrally.

Affected repos

• lex-apollo (persistence layer, would enforce/apply defaults)
• legion-apollo (client API, could accept scope from extension metadata)
• All extensions that write to Apollo (lex-microsoft_teams, lex-knowledge, legion-llm, etc.)