reverse-engineer: mark G6-G8 shipped + PROGRESS batch-4 (post-merge bookkeeping)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: Yoojin-nam

reverse_engineer/PROGRESS.md

@@ -34,3 +34,5 @@ One row per completed loop iteration. The loop appends here in Step E.
 - Next priority item (per PLAYBOOK): **batch 2 in progress — multi-skill rotation. Done: write-paper exemplar trio complete (iter 4 results + iter 5 discussion). Next: analyze-stats exemplar_tables, make-figures non-flow exemplar, check-reporting floor integration, a fresh-sector review probe.**
 
 - **Batch 3 — adjacent clinical-research scaffolds (2026-06-14 overnight, merged to main):** 4 PRs (#141 decision-curve+TRIPOD-LLM, #142 CONSORT-AI+SPIRIT-AI+RCT-AI probes, #143 diagnostic-accuracy/MRMC module+figure+table, #144 DECIDE-AI+decision-impact probes). G12–G23. Net: reporting guidelines 32→36 (TRIPOD-LLM, CONSORT-AI, SPIRIT-AI, DECIDE-AI), +1 domain-probe module (diagnostic_accuracy) + 4 probe subsections, +2 figure exemplars (decision_curve, mrmc_roc), +2 table-types (incremental_value, reader_study). All guideline-grounded educational summaries / synthetic anatomy under the license firewall; landed via a single conflict-reconciled integration merge.
+
+- **Batch 4 — backlog finish (2026-06-15, merged via #149):** G6 make-figures Bland–Altman + confusion-matrix exemplars (#146), G7 write-paper structured-abstract exemplar (#147), G8 check-reporting METRICS radiomics appraisal reference under `appraisal_tools/` — quality tool, NOT a counted reporting checklist, count stays 36 (#148). Ran as 2 parallel background lanes (G6/G7) + 1 direct (G8), reconciled into a single integration PR. Also #145: TRIPOD-LLM checklist renumbered to the official 19-item scheme after a fidelity verification pass (CONSORT-AI 14/14, SPIRIT-AI 15/15, DECIDE-AI 17/17 verified faithful). Scored gap backlog now exhausted.

reverse_engineer/gap_register.md

@@ -38,9 +38,9 @@ stop adding marginal items there).
 | G3 | make-figures | `exemplar_plots/roc_pr.md` — ROC / precision-recall anatomy (CI band, operating point, AUPRC under imbalance) | 4 | 5 | 5 | 100 | shipped (this PR) |
 | G4 | make-figures | `exemplar_plots/calibration_plot.md` — calibration anatomy (bins/loess, slope/intercept, distribution rug) | 4 | 4 | 5 | 80 | shipped (this PR) |
 | G5 | peer-review + self-review | `domain-probes/` RCT / intervention-trial probe (CONSORT: randomisation, allocation concealment, blinding, ITT, selective-outcome) — no trial probe despite trials being common | 5 | 4 | 5 | 100 | shipped (this PR) |
-| G6 | make-figures | `exemplar_plots/bland_altman.md` + `confusion_matrix.md` | 3 | 3 | 5 | 45 | in-progress (this PR) |
-| G7 | write-paper | `exemplar_introduction.md` + `exemplar_abstract.md` (the two sections without exemplars; section_guides exist) | 3 | 5 | 3 | 45 | intro shipped; abstract in-progress (this PR) |
-| G8 | check-reporting | `appraisal_tools/METRICS.md` — radiomics methodological-quality tool (named in the critical-item floor but no fuller reference). Placed under `appraisal_tools/` (NOT counted `checklists/`) because METRICS is a quality/RoB tool, not a reporting guideline — no reporting-guideline count change | 3 | 3 | 4 | 36 | in-progress (this PR) |
+| G6 | make-figures | `exemplar_plots/bland_altman.md` + `confusion_matrix.md` | 3 | 3 | 5 | 45 | shipped (#146) |
+| G7 | write-paper | `exemplar_introduction.md` + `exemplar_abstract.md` (the two sections without exemplars; section_guides exist) | 3 | 5 | 3 | 45 | shipped (#147) |
+| G8 | check-reporting | `appraisal_tools/METRICS.md` — radiomics methodological-quality tool (named in the critical-item floor but no fuller reference). Placed under `appraisal_tools/` (NOT counted `checklists/`) because METRICS is a quality/RoB tool, not a reporting guideline — no reporting-guideline count change | 3 | 3 | 4 | 36 | shipped (#148) |
 | G9 | calc-sample-size | `references/justification_examples.md` — reviewer-safe sample-size justification prose per design (found via cross-skill audit; SKILL.md promised "IRB-ready justification text" but no exemplar library) | 4 | 4 | 4 | 64 | shipped (this PR) |
 | G10 | present-paper | `scripts/inject_speaker_notes.py` run-level markdown parser — general speaker notes rendered `**bold**` literally (the failure mode pptx-speaker-notes.md warns against); the parser existed only in inject_pronunciation_notes.py. Found while triaging the unmerged `present-paper-md-notes-glossary` branch (whose verify_refs/academic-aio parts were already superseded by main) | 3 | 4 | 4 | 48 | shipped (rescue PR) |
 | G11 | manage-refs | `scripts/render_pandoc.sh` had no pre-render reference audit — a direct render call could ship fabricated/mismatched citations (the master pre_submission_gate audits, but direct calls bypass it). Found while triaging the unmerged present-paper-md-notes-glossary branch | 4 | 3 | 4 | 48 | shipped (cleanup PR) |