h3. The Ask

Today an answer in the playground transcript cites the same source twice: numbered badges on the figure card and a card per citation below it, each card carrying a provenance block. Replace both with a single ledger under the answer, and replace the status codes the transcript prints with plain-language lines.

Design, agreed on the canvas at [https://claude.ai/code/artifact/0a4e1e69-a662-40ec-958a-28bda029d20d] (page "Ledger", artboard "C · Ledger below"):

# The figure card holds only the answer and its qualifiers. No badges.
# Under it, one ledger with one row per source: number, page, what it is (table cell, chart value, text), the served value, a status dot. Clicking a row drives the document pane to the cited region as the reference cards do today. Row text is one colour throughout.
# On a flagged source only the served value and its dot turn red, and a plain-language line appears directly under that row, left-aligned, separated by a hairline, with room to read.
# A reconciled source shows no provenance block; value and status ride the row.
# When a cited cell's row line item is also cited, show the cell only.
# Everything else in the transcript (user bubbles, prose answers, the composer, batch cards, the document pane, highlight behaviour) renders as today.

*Status lines.* The transcript shows the registry label instead of the code. Wording for all fourteen codes in CELL_STATUS_REFERENCE, flags unchanged; only the six flagged codes print under a row:

* {{reconciled}}: Matches the page.
* {{header_printed_unlocated}}: This header is printed on the page. We could not measure where.
* {{label_printed_unlocated}}: This row label is printed on the page. We could not measure where.
* {{single_character}}: A one-character cell, such as a bare "$". Not matched to a position by design.
* {{total_label_added}}: The page prints no label on the totals row, so we wrote "Total".
* {{header_label_added}}: The page prints no header over this column, so we named it.
* {{header_text_dropped}}: The page prints header text here that we left out of the table.
* {{unverified}}: This value is on the page, but our check of it did not complete. Worth a manual look.
* {{defect}}: The page shows something different from what we recorded. See the note.
* {{value_misread}}: The page prints {printed} here, not {read}. Check which is right. (both figures from the flag's own record)
* {{value_unreconciled}}: Read from the chart once, without a second check. Worth a manual look.
* {{footnote_unresolved}}: This cell has a footnote marker, and we could not find the footnote it refers to.
* {{footnote_unreferenced}}: A footnote on this page that no marker points to.
* {{footnote_marker_unplaced}}: This table has a footnote marker we could not tie to a cell.

h3. Acceptance Criteria

* *AC1:* An answer with one figure cited in two places renders one figure card without badges and a two-row ledger beneath it; clicking a row highlights the cited region in the document pane.
* *AC2:* A flagged source renders its value and dot in red, its row text in the standard ink, and its plain-language line directly beneath the row; a reconciled source renders no line.
* *AC3:* The transcript never shows a status code; each of the fourteen codes renders its line above, with {{value_misread}} carrying both figures.
* *AC4:* A cell citation whose row line item is also cited renders once.
* *AC5:* Screenshots of the rendered transcript against a running playground, before and after, for a reconciled answer and a flagged answer.

*Out of scope:* changing which statuses are flagged (the inspect set), and changing the inspector's own notes.

h3. Current State

To Do. Design agreed on the canvas; implementation not started.

h3. Links

* Design canvas: [https://claude.ai/code/artifact/0a4e1e69-a662-40ec-958a-28bda029d20d]
* Predecessor: QUE-366 (merged, PR #209)
* Epic: QUE-92
* Standards: docs/ENGINEERING_STANDARDS.md