# Work queue census (extract, pinned at `95b6964`)

## What this is
A census table of every Q-section heading in the troupe work queue (`WORK-QUEUE.md`)
**as it stood at commit `95b69640feeb` (2026-10-28)**, the rev that wrote it.
One row per heading: the verbatim question label (`q`), the verbatim heading title,
the verbatim `**Status:**` line, and the verbatim first line of the `**Done means:**`
obligation.

A census is the class a subject held when it was read. `WORK-QUEUE.md` is the one
document every lane appends to, so this artifact names the rev it read and is graded
at that rev — not at whatever the file says today.

## Source
- `WORK-QUEUE.md` at `95b69640feeb0fa1c451bf84fd03895a38d6e27b`
  (blob `c431b2576f6fbf0f306a91be5e2740e8370c0cdc`, 96 `## Q` headings).
- Both are recorded in `manifest.json` (`rev`, `source_blob`) and graded by `CHECK`.

## Columns
| Column | Meaning |
|---|---|
| `q` | Question label, e.g. `Q1` |
| `title` | Heading title verbatim (leading `§k` sub-marks kept) |
| `status` | Verbatim `**Status:**` text |
| `done_implies` | First line of the `**Done means:**` obligation |

## Findings
- **96 rows** parsed from the source at the pin.
- **11 Q labels are re-opened** (the file is a running log and re-enters old items):
  `Q37, Q38, Q39, Q43, Q60, Q72, Q79, Q81, Q82, Q83, Q93`.
  These are **not** duplicates of distinct questions; they are the same label
  re-entered with a new title. 85 distinct labels in total.
- Status cells span multiple lines in the raw source; they are collapsed to single
  spaces so the TSV is one row per Q.

## Reproduce
  python3 extract_workqueue_census.py              # re-write at the pinned rev
  python3 extract_workqueue_census.py --rev <sha>  # re-pin and re-write

`extract_workqueue_census.py` is a wrapper over the verifier's `--write`, so the
builder and the checker cannot disagree about what a census is.

## Verify
  bash CHECK

`CHECK` runs `verify_workqueue_census.py --selftest` (13 cells) and then `--check`,
which grades 17 legs at the pin: the pin resolves, the queue blob at it matches the
manifest, the TSV re-derives **cell by cell in all four columns** in document order,
the markdown render matches the same derivation, the manifest's counts are measured,
and `extracted_at_utc` is not older than the pin's commit. It also prints a
`currency:` line saying how far HEAD has moved past the pin — advisory only, because
a census of a growing document is stale by construction.

Exit codes: `0` true at the pin, `1` a graded fault, `2` could not grade.