Explore a trace interactively¶

TL;DR

Generate qmstrace graph --out trace.html, open it, then search for the item being reviewed. Use its thread and finding inspector to locate source and review obligations.

From the product repo, generate a new snapshot:
```
qmstrace graph --out trace.html
```
Open trace.html in a browser. Read Evidence and limits before interpreting the graph. An unavailable source remains unavailable when a node is visible.
Search by ID or title and select an item from the list. Keyboard users can Tab to the matching button and press Enter. The inspector gives the full ID and source location; long labels may be shortened on the canvas. Hover over a node to see its full title without changing selection; untitled nodes show their full ID.
Set Direction to downstream to see what hangs off the item, or upstream to see its ancestry. Start with depth 1, then widen it. All items restores surrounding context.
Select an authored link in the inspector to read its marker and declaration location. Hidden links still contribute to the precomputed reach: filters change the picture, not the assessment. Check the hidden counts when interpreting an apparently isolated item.
In Findings, choose suspects to review S01. Read every root and governing tag, including historical-only reach. A missing current edge does not clear historical review work. Follow Review downstream suspects for the review itself.
After editing source, generate a new filename and reopen it. The old file is unchanged. Run qmstrace check --strict separately when a gate is needed.

The browser shows one repo, and its walks stop at that repo's edge. To walk an item across every repo of the project, run qmstrace ledger thread in the ledger.

The browser cannot record completion or approve a change. For a copyable change-request candidate use Write an impact list; for explicit test verdicts use Read test results.