How do I separate Go Montage transport errors from authorization limits?
A Go application can fail before it obtains any workspace evidence. Classify HTTP initialization, experimental auth negotiation and tool execution errors separately so downstream code does not mistake an unavailable authorized client for a workspace with no research coverage. The error artifact should identify the first unresolved layer rather than return a plausible empty workspace. Transport failure, experimental negotiation limits and rejected execution each require different remedies and must not be reported as zero research coverage.
Help me separate Go Montage transport errors from authorization limits. Review where the application turns SDK errors into its own result shape. Report the real workspace or the first unresolved failing stage, then complete supported cleanup. No writes.Locate error translation
Review where the application turns SDK errors into its own result shape. Identify whether failures from connection or auth are collapsed into an empty status object that callers might treat as a valid read.
Preserve the target and version
Keep the installed Streamable HTTP client and exact hosted endpoint during diagnosis. Switching URL suffixes or SDK examples without identifying the failed layer makes subsequent observations incomparable to the original failure.
Review auth-specific limitations
Compare an observed negotiation failure with the installed auth/oauthex documentation. Unsupported experimental behavior should remain blocked instead of triggering shared credentials, speculative enterprise assertions or an authorization downgrade.
Execute a direct read if qualified
Only with a supported approved auth path, invoke signals_status and inspect the actual tool result. A successful tool list provides metadata evidence but does not erase a later permission error returned during execution.
Return an honest result
Report the real workspace or the first unresolved failing stage, then complete supported cleanup. Missing authorization is not zero coverage, and a cleanly closed session is not evidence that the account-scoped read succeeded.
Keep the boundaries clear.
- Unsupported experimental behavior should remain blocked instead of triggering shared credentials, speculative enterprise assertions or an authorization downgrade.
- Missing authorization is not zero coverage, and a cleanly closed session is not evidence that the account-scoped read succeeded.
Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.
Primary references
Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.