m2 component-only mockup
Meaning-Layer Dashboard
This pass keeps the dashboard inside the existing component vocabulary so we can prove that the meaning-layer story does not depend on custom layout primitives. The page starts with what changed trust, where company context is exposed, and which unknowns still block fast response.
Needs review
3
+1
Boundary crossings
1
+1
Blind spots
2
Stable sessions
34
+4
Interpretation queue
What security should understand first
3 live stories
Story | Why it matters | Scope | Confidence |
|---|---|---|---|
Filesystem MCP server is the trust hinge | It bridges shared credentials, company files, and outbound model traffic. | Crosses user, endpoint, file, and external-path lanes. | Medium |
Shared key narrative is still unresolved | The key is known to be reused, but ownership and intended scope are not fully verified. | Affects several sessions and one sensitive repo lane. | Medium |
Two blind spots are delaying closure | Endpoint ownership and one supporting outbound host are still only partially explained. | Limits attribution quality more than it increases raw volume. | High |
Stable majority
Expected sessions
34
of 36 total
31
Known user + expected endpoint
+3
29
Expected model/provider lane
+2
34
No boundary crossing observed
+4
5
Contained low-signal MCP usage
+1
Blind spots
What still blocks confident closure
2 gaps
Endpoint owner missing
Weakens the trust floor for one active session.
Unblocks: Faster attribution on future session review.
Supporting host unlabeled
Makes one boundary crossing look broader than it may really be.
Unblocks: Cleaner classification of outbound traffic.