GeezerVison

Summary

• Remove workspace-scope-missing — it has no producer. The !context.workspaceId guard that emitted it disappeared when the copilot table tool's importRowsForModel was rewritten, and every operation now returns early on a missing workspace before provenance import is reachable. Its warn-classification test case constructed the reason itself, so it was asserting on something nothing emits.
• Give ResolvedSecretTraceProvenanceAccumulator the required reason the registry already has. Its three guards latched anonymously, which matters more here than on the registry: the wire format carries only complete, so the consumer can only ever say source-provenance-incomplete and the guard is unrecoverable. Names the three — file-source-unidentified, workspace-file-provenance-unknown, mcp-tool-execution-timeout. A latch from record() stays silent, since it reflects a bundle whose own registry already reported and subflow aggregation runs it per iteration.
• Fold the error/warn/by-design split into one reportIncompleteness. It was copied across both registry latches and would have been copied a third time for the accumulator; a copy that can be updated alone lets the same reason be a fault in one place and routine in another.
• Give the async workflow tool path its own import origin (copilotToolClient.workflowExecution) instead of latching with none, matching the sibling sealed-context path.
• Close UnrecordedDurableProvenanceCause, which was a free-form string carrying a TSDoc claim that it was always a static literal.

Also fixes a latent miss from #6513: the sweep passed a reason to the accumulator's zero-arg markIncomplete, but nothing caught it — test files are excluded from tsc, and the extra argument was ignored at runtime. That call is now real.

unspecified stays, with corrected docs. It has no production callsite by design, but tests use it as an honest "latched, no guard to name" lever; making them borrow a real reason would read as a claim about which guard tripped.

Type of Change

• Improvement (cleanup, no behavior change)

Testing

Verified every reason in the union now has at least one producer. bun run type-check clean, bun run lint clean, bun run check:audits — all 24 audits pass (including check:api-validation:strict). 3,140 tests green across the registry, durable-provenance, knowledge, providers, copilot-tools, logs, MCP, and file-manage suites. Two new tests pin the accumulator's naming and its silence on a recorded latch.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)