verify_task_body per-unit vocabulary scan: credit the what-is-plotted beat and strip-point phrasing
Highlight text on any card (body, plan, or an event) to anchor a comment, or leave a whole-task comment here. Mention @claude to summon a reply.
No comments yet.
target_file: scripts/verify_task_body.py candidate-fingerprint: perunit-vocab-scan-ignores-beat1-strip-point-phrasing
Goal
Fix verify_task_body.py's per-unit-evidence vocabulary scan (check-58 family): it WARNs on a result section whose what-is-plotted beat explicitly says the figure plots "every per-item mean as a strip point" — the scan does not credit the beat-1 line and its vocabulary list lacks the "strip point" phrasing, producing a false-positive WARN on a section that IS the per-unit view.
Evidence
Task #2477 clean-result round 2 (2026-08-23): the hero-strip result ("Base generation under the chat template fails the floor") carries beat-1 text "every per-item mean as a strip point" and the rendered per-item strips, yet the scan WARNed it as lacking per-unit evidence. Surfaced by the Claude clean-result-critic round-2 verdict (non-blocking observation, mechanizable: yes — extend the section-prose vocabulary window to include the beat-1 what-is-plotted line, or add "strip point" to the per-unit vocabulary list).
Acceptance
- A section whose what-is-plotted beat contains per-unit vocabulary (incl. "strip point") no longer WARNs.
- Regression test reproducing the #2477 hero-section shape.