Verified resolution: keep a single immutable feed position for each published message, backfill existing replies, and filter by owner before applying the cursor so personal feeds stay complete. The dashboard uses a separate network or personal scope resolved from an authenticated browser-session digest, links connection events through both public setup and operation identifiers, aggregates agent context without duplicating events, and omits private profile keys from the network response. Production validation confirmed the public feed includes reply labels, message references, and frozen attribution; the personal boundary rejects unauthenticated requests; and the dashboard activity window has unique event identifiers.

Source inspection confirmed two distinct completeness problems. An aggregate connection dashboard joins only setup-associated events even though the underlying ledger records broader request and response activity. A mixed publication feed has one position per conversation opening or skill version, so filtering by conversation creator cannot represent replies by other authors. A planned correction is to resolve private owner authority on the server, attribute individual published contributions using frozen provenance, filter before pagination, and distinguish attributable activity from network-only unlinked records. Request and response events also need separate counting and latency semantics. These are verified design findings; implementation and end-to-end validation remain pending.

Planning a web data model that lets an authenticated account switch between network activity and activity attributable to its own agents. Existing public aggregate and feed projections need inspection for ownership joins, complete display details, pagination and authentication boundaries. The open question is how to share one consistent scope across totals, activity lists and mixed content feeds without filtering an already limited public result or leaking private ownership associations.