Release v1.6.0: fix spacing detection, add page refs and engine footer

- Fix FMT-009 false positive: measure inter-block baseline gaps for PDFs
  that encode each visual line as a separate text block (was only measuring
  intra-block gaps, missing double spacing entirely on most briefs)
- Add per-check page numbers to HTML report (badges on failed checks,
  column on passed checks table)
- Add PyMuPDF/fallback status line to report footer
- Add pages field to CheckResult, pymupdf_used field to ComplianceReport
- Add --pymupdf/--no-pymupdf CLI flags to build_report.py
- Update known issues: FMT-009 removed from active false-positive list

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: jet52

SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: brief-compliance
-version: 1.5.0
+version: 1.6.0
 description: >-
   Triggers when a user uploads a legal brief PDF for compliance review against the
   North Dakota Rules of Appellate Procedure. Analyzes the brief and produces a
@@ -175,12 +175,13 @@ Based on testing across 13 briefs (Feb 2026):
 
 ### Mechanical Check False Positives
 
-Three mechanical checks have high false-positive rates. When reporting results, note these caveats to the user:
+Two mechanical checks have high false-positive rates. When reporting results, note these caveats to the user:
 
 - **FMT-006 (Font Size)**: Measures the minimum font found anywhere in the PDF. Small fonts in page numbers, headers, footers, superscripts, or PDF artifacts trigger REJECT even when the body text is properly 12pt. If this is the sole REJECT trigger and the reported minimum is 8-11pt, flag it as a likely false positive.
-- **FMT-009 (Spacing)**: Nearly all briefs are flagged as single-spaced. The detector is miscalibrated for many PDF encodings. If the brief appears to be a standard attorney-prepared document, note this is likely a false positive.
 - **FMT-005 (Bottom Margin)**: Page numbers at the bottom are measured as content in the margin zone. Nearly always triggers.
 
+**Fixed (v1.6.0):** FMT-009 (Spacing) previously only measured intra-block line gaps, missing PDFs that encode each line as a separate text block. Now also measures inter-block baseline distances, which correctly detects double spacing in these PDFs.
+
 ### Brief Type Auto-Detection
 
 The `--brief-type auto` flag frequently returns "unknown", especially for appellee briefs. If auto-detection fails, re-run Phase 1 with an explicit `--brief-type` flag based on the cover page text.
@@ -224,7 +225,7 @@ These are run by `check_brief.py` — no changes needed here.
 | FMT-006 | Font size >= 12pt | 32(a)(5) | REJECT | ⚠ High false-positive rate — flags small fonts in page numbers, headers, superscripts |
 | FMT-007 | Max 16 chars/inch | 32(a)(5) | CORRECTION |
 | FMT-008 | Plain roman style | 32(a)(6) | NOTE |
-| FMT-009 | Double-spaced body text | 32(a)(5) | CORRECTION | ⚠ High false-positive rate — spacing detection miscalibrated for many PDF encodings |
+| FMT-009 | Double-spaced body text | 32(a)(5) | CORRECTION |
 | FMT-010 | Footnotes double-spaced, same typeface | 32(a)(5) | NOTE |
 | FMT-011 | Pages numbered at bottom | 32(a)(4) | CORRECTION |
 | FMT-012 | Numbering starts with "1" on cover | 32(a)(4) | NOTE |

core/models.py

@@ -39,6 +39,7 @@ class CheckResult:
     message: str
     details: Optional[str] = None
     applicable: bool = True  # False if this check doesn't apply to the brief type
+    pages: Optional[list[int]] = None  # Page numbers with issues (if applicable)
 
     @property
     def failed(self) -> bool:
@@ -91,6 +92,7 @@ class ComplianceReport:
     case_title: str = ""        # e.g. "Rath v. Rath et al."
     brief_label: str = ""       # e.g. "Amended Brief of Defendant-Appellant"
     pdf_filename: str = ""      # original PDF filename
+    pymupdf_used: bool = True   # Whether PyMuPDF was used for mechanical checks
 
     @property
     def failed_checks(self) -> list[CheckResult]:

core/pdf_extract.py

@@ -200,23 +200,44 @@ def _compute_margins(blocks: list[dict], rect: fitz.Rect) -> tuple[float, float,
 
 
 def _estimate_line_spacing(blocks: list[dict]) -> Optional[float]:
-    """Estimate typical line spacing in points from text block baselines."""
+    """Estimate typical line spacing in points from text block baselines.
+
+    Measures both intra-block line gaps and inter-block gaps (for PDFs that
+    encode each visual line as a separate block).
+    """
     spacings = []
+
+    # 1. Intra-block: gaps between lines within the same block
     for block in blocks:
         if block["type"] != 0:
             continue
         lines = block.get("lines", [])
         for i in range(1, len(lines)):
-            prev_baseline = lines[i - 1]["bbox"][3]  # bottom of previous line
-            curr_baseline = lines[i]["bbox"][1]  # top of current line
-            # More accurate: use the origin y of spans
             prev_origins = [s["origin"][1] for s in lines[i - 1].get("spans", []) if s["text"].strip()]
             curr_origins = [s["origin"][1] for s in lines[i].get("spans", []) if s["text"].strip()]
             if prev_origins and curr_origins:
                 spacing = min(curr_origins) - min(prev_origins)
-                if 8 < spacing < 60:  # reasonable range
+                if 8 < spacing < 60:
                     spacings.append(spacing)
 
+    # 2. Inter-block: gaps between consecutive single-line text blocks.
+    #    Many PDF generators emit each line as its own block, so intra-block
+    #    measurement finds nothing. Walk consecutive text blocks and measure
+    #    baseline-to-baseline distance.
+    text_blocks = [b for b in blocks if b.get("type") == 0]
+    for i in range(1, len(text_blocks)):
+        prev_lines = text_blocks[i - 1].get("lines", [])
+        curr_lines = text_blocks[i].get("lines", [])
+        if not prev_lines or not curr_lines:
+            continue
+        # Use the last line of prev block and first line of curr block
+        prev_origins = [s["origin"][1] for s in prev_lines[-1].get("spans", []) if s["text"].strip()]
+        curr_origins = [s["origin"][1] for s in curr_lines[0].get("spans", []) if s["text"].strip()]
+        if prev_origins and curr_origins:
+            spacing = min(curr_origins) - min(prev_origins)
+            if 8 < spacing < 60:
+                spacings.append(spacing)
+
     return statistics.median(spacings) if spacings else None
 
 

core/report_builder.py

@@ -109,6 +109,7 @@ def build_html_report(report: ComplianceReport, version_stamp: str = "") -> str:
 
   <footer>
     <p>Generated by the ND Supreme Court Brief Compliance Checker{f' &middot; {_esc(version_stamp)}' if version_stamp else ''}</p>
+    <p class="engine-note">Mechanical checks: {'PyMuPDF' if report.pymupdf_used else 'Fallback (PyMuPDF unavailable — mechanical checks skipped)'}</p>
   </footer>
 </div>
 </body>
@@ -121,6 +122,10 @@ def _render_check_group(checks: list[CheckResult], title: str, color: str) -> st
     rows = ""
     for c in checks:
         detail = f'<div class="detail">{_esc(c.details)}</div>' if c.details else ""
+        pages_html = ""
+        if c.pages:
+            page_tags = " ".join(f'<span class="page-badge">p.{p}</span>' for p in c.pages)
+            pages_html = f'<div class="check-pages">Pages: {page_tags}</div>'
         rows += f"""<div class="check-card failed" style="border-left-color:{color};">
   <div class="check-header">
     <span class="check-id">{c.check_id}</span>
@@ -129,6 +134,7 @@ def _render_check_group(checks: list[CheckResult], title: str, color: str) -> st
     <span class="severity severity-{c.severity.value}">{c.severity.value.upper()}</span>
   </div>
   <div class="check-msg">{_esc(c.message)}</div>
+  {pages_html}
   {detail}
 </div>
 """
@@ -140,15 +146,19 @@ def _render_checks_table(checks: list[CheckResult], passed: bool = True) -> str:
         return ""
     rows = ""
     for c in checks:
+        pages_cell = ""
+        if c.pages:
+            pages_cell = ", ".join(str(p) for p in c.pages)
         rows += f"""<tr>
   <td class="check-id-cell">{c.check_id}</td>
   <td>{_esc(c.name)}</td>
   <td>{_rule_link(c.rule)}</td>
   <td>{_esc(c.message)}</td>
+  <td class="pages-cell">{pages_cell}</td>
 </tr>
 """
     return f"""<table class="checks-table">
-<thead><tr><th>ID</th><th>Check</th><th>Rule</th><th>Result</th></tr></thead>
+<thead><tr><th>ID</th><th>Check</th><th>Rule</th><th>Result</th><th>Pages</th></tr></thead>
 <tbody>{rows}</tbody>
 </table>"""
 
@@ -228,7 +238,11 @@ def _css() -> str:
 details summary::-webkit-details-marker { display: none; }
 details summary::before { content: '▶ '; font-size: 0.8rem; }
 details[open] summary::before { content: '▼ '; }
+.check-pages { margin-top: 0.3rem; font-size: 0.85rem; color: #586069; }
+.page-badge { display: inline-block; background: #e1e4e8; color: #24292e; font-size: 0.78rem; font-family: monospace; padding: 0.1rem 0.35rem; border-radius: 3px; margin: 0 0.15rem; }
+.pages-cell { font-family: monospace; font-size: 0.85rem; color: #586069; white-space: nowrap; }
 footer { text-align: center; color: #586069; font-size: 0.85rem; padding: 1.5rem 0; border-top: 1px solid #e1e4e8; margin-top: 2rem; }
+.engine-note { font-size: 0.8rem; margin-top: 0.25rem; }
 @media print { .container { max-width: 100%; } .banner { border-width: 3px; } details { open: true; } details[open] summary { display: none; } }
 @media (max-width: 600px) { .check-header { flex-direction: column; align-items: flex-start; } .severity { margin-left: 0; } }
 """

scripts/build_report.py

@@ -27,10 +27,33 @@
 from core.version_check import get_version_stamp
 
 
+def _extract_pages_from_message(message: str) -> list[int] | None:
+    """Try to extract page numbers from a check message like 'Top Margin < 1" on 2, 3, 4'."""
+    # Match patterns like "on 2, 3, 4, 5" or "on pages 2, 3, 4"
+    m = re.search(r'\bon(?:\s+pages?)?\s+([\d,\s]+)', message)
+    if m:
+        nums = re.findall(r'\d+', m.group(1))
+        if nums:
+            return [int(n) for n in nums]
+    # Match "pages: 2, 3" in details
+    m = re.search(r'pages?[:\s]+([\d,\s]+)', message, re.IGNORECASE)
+    if m:
+        nums = re.findall(r'\d+', m.group(1))
+        if nums:
+            return [int(n) for n in nums]
+    return None
+
+
 def _parse_results(items: list[dict]) -> list[CheckResult]:
     """Convert a list of JSON dicts into CheckResult objects."""
     results = []
     for item in items:
+        pages = item.get("pages")
+        if pages is None and not item["passed"]:
+            # Try to extract page numbers from message or details
+            pages = _extract_pages_from_message(item["message"])
+            if pages is None and item.get("details"):
+                pages = _extract_pages_from_message(item["details"])
         results.append(CheckResult(
             check_id=item["check_id"],
             name=item["name"],
@@ -40,6 +63,7 @@ def _parse_results(items: list[dict]) -> list[CheckResult]:
             message=item["message"],
             details=item.get("details"),
             applicable=item.get("applicable", True),
+            pages=pages,
         ))
     return results
 
@@ -130,6 +154,8 @@ def main():
     parser.add_argument("--semantic", required=True, help="Path to semantic results JSON from Claude Code")
     parser.add_argument("--output-dir", default=None, help="Directory for HTML report (default: same as intermediate)")
     parser.add_argument("--reasoning", default=None, help="Optional reasoning text for the report summary")
+    parser.add_argument("--pymupdf", action="store_true", default=True, help="PyMuPDF was used for mechanical checks (default)")
+    parser.add_argument("--no-pymupdf", action="store_false", dest="pymupdf", help="PyMuPDF was NOT used (fallback mode)")
     args = parser.parse_args()
 
     intermediate_path = Path(args.intermediate)
@@ -186,6 +212,7 @@ def main():
         case_title=case_title,
         brief_label=brief_label,
         pdf_filename=Path(pdf_path_str).name,
+        pymupdf_used=args.pymupdf,
     )
 
     html = build_html_report(report, version_stamp=get_version_stamp())

version.json

@@ -1,6 +1,6 @@
 {
-  "version": "1.5.0",
-  "build_date": "2026-02-20",
+  "version": "1.6.0",
+  "build_date": "2026-02-25",
   "rules_verified": "2026-02-17",
   "rules_freshness_days": 90,
   "check_url": "https://raw.githubusercontent.com/jet52/ndsc-brief-compliance/main/version.json",