Apply the #1392 runbook fixes to the Aider and Devin monolith skills

The monoliths are hand-maintained single files frozen against a pinned pristine-v8 blob by the round-trip guard, so they were excluded from the 0.8.44 #1392 batch. Evolve the guard from a positional zip (line-count-exact, single-line-class allowlist) to a multiset diff that classifies every added/removed line against documented sanctioned change-classes, so the multi-line fixes can land while any unsanctioned drift still fails. Add predicates for the four fix classes and broaden the enum/chunk-cleanup predicates to match both the v8 and rewritten forms.

Both monoliths now: thread directed=IS_DIRECTED through every build_from_json call (a --directed run no longer collapses reciprocal edges), scope semantic extraction to document/paper/image, unlink a stale .graphify_cached.json on a cache miss, and run Step 4's zero-node guard before any write with the report/analysis gated on to_json persisting the graph (#1392).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

CHANGELOG.md

@@ -4,6 +4,8 @@ Full release notes with details on each version: [GitHub Releases](https://githu
 
 ## Unreleased
 
+- Fix: the Aider and Devin monolith skills now carry the #1392 runbook fixes that the split skill got in 0.8.44. These single-file skills are hand-maintained and frozen against a pinned pristine-v8 blob by a round-trip guard, so they had been excluded. The guard is now a multiset diff that classifies every added/removed line against documented sanctioned change-classes (rather than a positional zip that forbade any line-count change), which lets the multi-line fixes land while still failing on any unsanctioned drift. Both monoliths now propagate `directed=IS_DIRECTED` into every `build_from_json` call (a `--directed` run no longer collapses reciprocal edges), scope semantic extraction to document/paper/image (code is covered by the AST pass), delete `.graphify_cached.json` on a cache miss, and run Step 4's zero-node guard before any write with the report/analysis gated on `to_json` actually persisting the graph (#1392).
+
 ## 0.8.44 (2026-06-19)
 
 - Fix: generated Claude/agent skill, crash & data-loss bugs in the runbooks (#1392). (1) Semantic chunk files were written under the **scanned dir** (`.graphify_root`) but the merge globs **cwd** `graphify-out/`, so a non-cwd scan produced "no nodes"; chunk paths are now derived from cwd. (2) Code-only corpora skipped Part B but Part C reads `.graphify_semantic.json` unconditionally, raising `FileNotFoundError`; the fast path now writes an empty semantic file first. (3) `--cluster-only` told the agent to re-run Steps 5-9, which read intermediate files a prior cleanup deleted (`FileNotFoundError`); it now relies on the self-contained `graphify cluster-only` CLI. (4) Step 4's zero-node guard ran *after* `GRAPH_REPORT.md`/`graph.json`/analysis were written, and `GRAPH_REPORT.md` was written before `to_json`'s #479 shrink-guard; the guard now runs before any write and the report/analysis are written only when `to_json` actually persisted the graph.

graphify/skill-aider.md

@@ -226,12 +226,19 @@ from graphify.cache import check_semantic_cache
 from pathlib import Path
 
 detect = json.loads(Path('.graphify_detect.json').read_text())
-all_files = [f for files in detect['files'].values() for f in files]
+# Only content files go to semantic extraction. Code is already covered
+# structurally by the AST pass; flattening every category here makes the
+# extraction step re-read every source file (#1392).
+all_files = [f for cat in ('document', 'paper', 'image') for f in detect['files'].get(cat, [])]
 
 cached_nodes, cached_edges, cached_hyperedges, uncached = check_semantic_cache(all_files)
 
+# Always (re)write the cache file: write hits, else DELETE any leftover from a
+# prior run so Part C never merges a stale .graphify_cached.json (#1392).
 if cached_nodes or cached_edges or cached_hyperedges:
     Path('.graphify_cached.json').write_text(json.dumps({'nodes': cached_nodes, 'edges': cached_edges, 'hyperedges': cached_hyperedges}))
+else:
+    Path('.graphify_cached.json').unlink(missing_ok=True)
 Path('.graphify_uncached.txt').write_text('\n'.join(uncached))
 print(f'Cache: {len(all_files)-len(uncached)} files hit, {len(uncached)} files need extraction')
 "
@@ -377,6 +384,8 @@ print(f'Merged: {total} nodes, {edges} edges ({len(ast[\"nodes\"])} AST + {len(s
 
 ### Step 4 - Build graph, cluster, analyze, generate outputs
 
+**Before starting:** the code blocks below pass `directed=IS_DIRECTED` to `build_from_json()`. Replace `IS_DIRECTED` with `True` if `--directed` was given (builds a `DiGraph` preserving edge direction source->target), otherwise `False` (the default undirected `Graph`). Substitute it everywhere it appears, the same way you substitute `INPUT_PATH` - do not leave the literal `IS_DIRECTED` in the code.
+
 ```bash
 mkdir -p graphify-out
 $(cat graphify-out/.graphify_python) -c "
@@ -391,7 +400,13 @@ from pathlib import Path
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 detection  = json.loads(Path('.graphify_detect.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
+# Guard BEFORE any write: an empty extraction must not clobber a good graph.json /
+# GRAPH_REPORT.md / analysis sidecar. Check immediately after build (#1392).
+if G.number_of_nodes() == 0:
+    print('ERROR: Graph is empty - extraction produced no nodes.')
+    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
+    raise SystemExit(1)
 communities = cluster(G)
 cohesion = score_all(G, communities)
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -401,9 +416,15 @@ labels = {cid: 'Community ' + str(cid) for cid in communities}
 # Placeholder questions - regenerated with real labels in Step 5
 questions = suggest_questions(G, communities, labels)
 
+# Persist the graph first and only write the report/analysis if it actually
+# persisted - to_json refuses to shrink an existing graph.json (#479), and a
+# report describing a graph we did not write would be a lie (#1392).
+wrote = to_json(G, communities, 'graphify-out/graph.json')
+if not wrote:
+    print('ERROR: refused to shrink graphify-out/graph.json (fewer nodes than the existing graph). Run a full rebuild to be safe.')
+    raise SystemExit(1)
 report = generate(G, communities, cohesion, labels, gods, surprises, detection, tokens, 'INPUT_PATH', suggested_questions=questions)
 Path('graphify-out/GRAPH_REPORT.md').write_text(report)
-to_json(G, communities, 'graphify-out/graph.json')
 
 analysis = {
     'communities': {str(k): v for k, v in communities.items()},
@@ -413,10 +434,6 @@ analysis = {
     'questions': questions,
 }
 Path('.graphify_analysis.json').write_text(json.dumps(analysis, indent=2))
-if G.number_of_nodes() == 0:
-    print('ERROR: Graph is empty - extraction produced no nodes.')
-    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
-    raise SystemExit(1)
 print(f'Graph: {G.number_of_nodes()} nodes, {G.number_of_edges()} edges, {len(communities)} communities')
 "
 ```
@@ -444,7 +461,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 detection  = json.loads(Path('.graphify_detect.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -482,7 +499,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -513,7 +530,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -536,7 +553,7 @@ from graphify.build import build_from_json
 from graphify.export import to_cypher
 from pathlib import Path
 
-G = build_from_json(json.loads(Path('.graphify_extract.json').read_text()))
+G = build_from_json(json.loads(Path('.graphify_extract.json').read_text()), directed=IS_DIRECTED)
 to_cypher(G, 'graphify-out/cypher.txt')
 print('cypher.txt written - import with: cypher-shell < graphify-out/cypher.txt')
 "
@@ -554,7 +571,7 @@ from pathlib import Path
 
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 result = push_to_neo4j(G, uri='NEO4J_URI', user='NEO4J_USER', password='NEO4J_PASSWORD', communities=communities)
@@ -577,7 +594,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -598,7 +615,7 @@ from pathlib import Path
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 to_graphml(G, communities, 'graphify-out/graph.graphml')
@@ -795,7 +812,7 @@ G_existing = json_graph.node_link_graph(existing_data, edges='links')
 
 # Load new extraction
 new_extraction = json.loads(Path('.graphify_extract.json').read_text())
-G_new = build_from_json(new_extraction)
+G_new = build_from_json(new_extraction, directed=IS_DIRECTED)
 
 # Merge: new nodes/edges into existing graph
 G_existing.update(G_new)
@@ -819,7 +836,7 @@ from pathlib import Path
 # Load old graph (before update) from backup written before merge
 old_data = json.loads(Path('.graphify_old.json').read_text()) if Path('.graphify_old.json').exists() else None
 new_extract = json.loads(Path('.graphify_extract.json').read_text())
-G_new = build_from_json(new_extract)
+G_new = build_from_json(new_extract, directed=IS_DIRECTED)
 
 if old_data:
     G_old = json_graph.node_link_graph(old_data, edges='links')

graphify/skill-devin.md

@@ -243,12 +243,19 @@ from graphify.cache import check_semantic_cache
 from pathlib import Path
 
 detect = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
-all_files = [f for files in detect['files'].values() for f in files]
+# Only content files go to semantic extraction. Code is already covered
+# structurally by the AST pass; flattening every category here makes the
+# extraction step re-read every source file (#1392).
+all_files = [f for cat in ('document', 'paper', 'image') for f in detect['files'].get(cat, [])]
 
 cached_nodes, cached_edges, cached_hyperedges, uncached = check_semantic_cache(all_files)
 
+# Always (re)write the cache file: write hits, else DELETE any leftover from a
+# prior run so Part C never merges a stale .graphify_cached.json (#1392).
 if cached_nodes or cached_edges or cached_hyperedges:
     Path('graphify-out/.graphify_cached.json').write_text(json.dumps({'nodes': cached_nodes, 'edges': cached_edges, 'hyperedges': cached_hyperedges}))
+else:
+    Path('graphify-out/.graphify_cached.json').unlink(missing_ok=True)
 Path('graphify-out/.graphify_uncached.txt').write_text('\n'.join(uncached))
 print(f'Cache: {len(all_files)-len(uncached)} files hit, {len(uncached)} files need extraction')
 "
@@ -442,6 +449,8 @@ print(f'Merged: {total} nodes, {edges} edges ({len(ast[\"nodes\"])} AST + {len(s
 
 ### Step 4 - Build graph, cluster, analyze, generate outputs
 
+**Before starting:** the code blocks below pass `directed=IS_DIRECTED` to `build_from_json()`. Replace `IS_DIRECTED` with `True` if `--directed` was given (builds a `DiGraph` preserving edge direction source->target), otherwise `False` (the default undirected `Graph`). Substitute it everywhere it appears, the same way you substitute `INPUT_PATH` - do not leave the literal `IS_DIRECTED` in the code.
+
 ```bash
 mkdir -p graphify-out
 $(cat graphify-out/.graphify_python) -c "
@@ -456,7 +465,13 @@ from pathlib import Path
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 detection  = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
+# Guard BEFORE any write: an empty extraction must not clobber a good graph.json /
+# GRAPH_REPORT.md / analysis sidecar. Check immediately after build (#1392).
+if G.number_of_nodes() == 0:
+    print('ERROR: Graph is empty - extraction produced no nodes.')
+    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
+    raise SystemExit(1)
 communities = cluster(G)
 cohesion = score_all(G, communities)
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -466,9 +481,15 @@ labels = {cid: 'Community ' + str(cid) for cid in communities}
 # Placeholder questions - regenerated with real labels in Step 5
 questions = suggest_questions(G, communities, labels)
 
+# Persist the graph first and only write the report/analysis if it actually
+# persisted - to_json refuses to shrink an existing graph.json (#479), and a
+# report describing a graph we did not write would be a lie (#1392).
+wrote = to_json(G, communities, 'graphify-out/graph.json')
+if not wrote:
+    print('ERROR: refused to shrink graphify-out/graph.json (fewer nodes than the existing graph). Run a full rebuild to be safe.')
+    raise SystemExit(1)
 report = generate(G, communities, cohesion, labels, gods, surprises, detection, tokens, 'INPUT_PATH', suggested_questions=questions)
 Path('graphify-out/GRAPH_REPORT.md').write_text(report)
-to_json(G, communities, 'graphify-out/graph.json')
 
 analysis = {
     'communities': {str(k): v for k, v in communities.items()},
@@ -478,10 +499,6 @@ analysis = {
     'questions': questions,
 }
 Path('graphify-out/.graphify_analysis.json').write_text(json.dumps(analysis, indent=2))
-if G.number_of_nodes() == 0:
-    print('ERROR: Graph is empty - extraction produced no nodes.')
-    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
-    raise SystemExit(1)
 print(f'Graph: {G.number_of_nodes()} nodes, {G.number_of_edges()} edges, {len(communities)} communities')
 "
 ```
@@ -509,7 +526,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 detection  = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -547,7 +564,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -578,7 +595,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -632,7 +649,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -655,7 +672,7 @@ from graphify.build import build_from_json
 from graphify.export import to_cypher
 from pathlib import Path
 
-G = build_from_json(json.loads(Path('graphify-out/.graphify_extract.json').read_text()))
+G = build_from_json(json.loads(Path('graphify-out/.graphify_extract.json').read_text()), directed=IS_DIRECTED)
 to_cypher(G, 'graphify-out/cypher.txt')
 print('cypher.txt written - import with: cypher-shell < graphify-out/cypher.txt')
 "
@@ -672,7 +689,7 @@ from pathlib import Path
 
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 result = push_to_neo4j(G, uri='NEO4J_URI', user='NEO4J_USER', password='NEO4J_PASSWORD', communities=communities)
@@ -695,7 +712,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -716,7 +733,7 @@ from pathlib import Path
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 to_graphml(G, communities, 'graphify-out/graph.graphml')
@@ -932,7 +949,7 @@ G_existing = json_graph.node_link_graph(existing_data, edges='links')
 
 # Load new extraction
 new_extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
-G_new = build_from_json(new_extraction)
+G_new = build_from_json(new_extraction, directed=IS_DIRECTED)
 
 # Merge: new nodes/edges into existing graph
 G_existing.update(G_new)
@@ -955,7 +972,7 @@ from pathlib import Path
 
 old_data = json.loads(Path('graphify-out/.graphify_old.json').read_text()) if Path('graphify-out/.graphify_old.json').exists() else None
 new_extract = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
-G_new = build_from_json(new_extract)
+G_new = build_from_json(new_extract, directed=IS_DIRECTED)
 
 if old_data:
     G_old = json_graph.node_link_graph(old_data, edges='links')

tests/test_skillgen.py

@@ -450,54 +450,57 @@ def test_monolith_roundtrip_passes_for_aider_and_devin():
         assert problems == [], f"[{key}]\n" + "\n".join(problems)
 
 
-def test_monoliths_change_only_the_enum_description_and_chunk_cleanup():
-    """The rendered monolith differs from v8 on exactly the allowed lines.
+def test_monoliths_change_only_sanctioned_lines():
+    """Every line that differs from pristine v8 is a sanctioned change-class.
 
-    Three changes are now in play for the monoliths: the file_type enum unified to
-    the six-value superset (the prose guidance line + the schema line), the
-    frontmatter description unified across all platforms, and the shell-agnostic
-    chunk-cleanup rewrite (#1172). Nothing else may differ.
+    The round-trip (multiset diff vs the pinned v8 blob) must come back clean:
+    each added/removed line matches one of the documented sanctioned predicates
+    in gen — the enum unification, the unified description, the chunk-cleanup
+    rewrite (#1172), and the four #1392 runbook fixes. Anything else is drift.
     """
     platforms = gen.load_platforms()
     for key in ("aider", "devin"):
-        rendered = gen.render(platforms[key])[0].content.splitlines()
-        # Strip trigger: lines from the reference — their removal (#1180) is a
-        # permitted diff alongside enum, description, and chunk-cleanup changes.
-        original = [
-            l for l in gen._normalise(gen._git_show(platforms[key].roundtrip_ref)).splitlines()
-            if not gen._is_trigger_line(l)
-        ]
-        assert len(rendered) == len(original), f"[{key}] line count changed"
-        diff_idx = [i for i, (r, o) in enumerate(zip(rendered, original)) if r != o]
-        # Four lines change: the prose enum guidance, the schema line,
-        # the frontmatter description, and the chunk-cleanup rewrite.
-        assert len(diff_idx) == 4, f"[{key}] expected 4 changed lines, got {len(diff_idx)}"
-        enum_changes = 0
-        desc_changes = 0
-        cleanup_changes = 0
-        for i in diff_idx:
-            line = rendered[i]
-            if gen.ENUM_VALUES in line or gen.ENUM_PROSE in line:
-                enum_changes += 1
-            elif line.lstrip().startswith("description:"):
-                desc_changes += 1
-                assert UNIFIED_DESCRIPTION in line, (
-                    f"[{key}] description line is not the unified text: {line!r}"
-                )
-            elif gen._is_chunk_cleanup_line(line):
-                cleanup_changes += 1
-                # The unmatched-glob abort is fixed: the rm no longer carries the
-                # bare chunk glob, and a find ... -delete sweeps the chunks.
-                assert ".graphify_chunk_*.json" not in line.split("find", 1)[0]
-            else:
-                raise AssertionError(
-                    f"[{key}] changed line {i} is none of enum/description/cleanup: {line!r}"
-                )
-        assert enum_changes == 2, f"[{key}] expected 2 enum line changes, got {enum_changes}"
-        assert desc_changes == 1, f"[{key}] expected 1 description change, got {desc_changes}"
-        assert cleanup_changes == 1, f"[{key}] expected 1 cleanup change, got {cleanup_changes}"
+        assert gen.monolith_roundtrip(platforms[key]) == []
         # The six-value superset replaced the five-value enum in both files.
-        assert any(gen.ENUM_VALUES in line for line in rendered)
+        rendered = gen.render(platforms[key])[0].content
+        assert gen.ENUM_VALUES in rendered
+        assert UNIFIED_DESCRIPTION in rendered
+
+
+def test_monoliths_carry_the_1392_runbook_fixes():
+    """The four #1392 data-loss/correctness fixes are present in both monoliths.
+
+    The round-trip allows these change-classes; this test asserts they are
+    actually applied, so a regression that drops a fix fails here even though the
+    round-trip (which only forbids *unsanctioned* drift) would still pass.
+    """
+    platforms = gen.load_platforms()
+    for key in ("aider", "devin"):
+        body = gen.render(platforms[key])[0].content
+
+        # #6/#7 directed propagation: no bare build_from_json call survives, and
+        # the IS_DIRECTED substitution instruction is present.
+        assert "directed=IS_DIRECTED" in body
+        assert "build_from_json(extraction)" not in body
+        assert "Substitute it everywhere it appears" in body
+
+        # #10 content-only semantic scope: code is no longer flattened in.
+        assert "for cat in ('document', 'paper', 'image')" in body
+        assert "detect['files'].values()" not in body
+
+        # #12 stale-cache unlink on a miss.
+        assert ".graphify_cached.json').unlink(missing_ok=True)" in body
+
+        # #18/#20 zero-node guard before any write, report/analysis gated on
+        # to_json's return.
+        lines = body.splitlines()
+        build_i = next(i for i, l in enumerate(lines) if "G = build_from_json(extraction, directed=IS_DIRECTED)" in l)
+        guard_i = next(i for i, l in enumerate(lines[build_i:], build_i) if "number_of_nodes() == 0" in l)
+        report_i = next(i for i, l in enumerate(lines[build_i:], build_i) if "GRAPH_REPORT.md').write_text(report)" in l)
+        wrote_i = next(i for i, l in enumerate(lines[build_i:], build_i) if l.strip().startswith("wrote = to_json("))
+        # guard fires right after the build, before the graph/report are written.
+        assert build_i < guard_i < wrote_i < report_i, f"[{key}] Step 4 ordering not fixed"
+        assert "if not wrote:" in body
 
 
 def test_devin_keeps_its_multi_field_frontmatter():

tools/skillgen/expected/graphify__skill-aider.md

@@ -226,12 +226,19 @@ from graphify.cache import check_semantic_cache
 from pathlib import Path
 
 detect = json.loads(Path('.graphify_detect.json').read_text())
-all_files = [f for files in detect['files'].values() for f in files]
+# Only content files go to semantic extraction. Code is already covered
+# structurally by the AST pass; flattening every category here makes the
+# extraction step re-read every source file (#1392).
+all_files = [f for cat in ('document', 'paper', 'image') for f in detect['files'].get(cat, [])]
 
 cached_nodes, cached_edges, cached_hyperedges, uncached = check_semantic_cache(all_files)
 
+# Always (re)write the cache file: write hits, else DELETE any leftover from a
+# prior run so Part C never merges a stale .graphify_cached.json (#1392).
 if cached_nodes or cached_edges or cached_hyperedges:
     Path('.graphify_cached.json').write_text(json.dumps({'nodes': cached_nodes, 'edges': cached_edges, 'hyperedges': cached_hyperedges}))
+else:
+    Path('.graphify_cached.json').unlink(missing_ok=True)
 Path('.graphify_uncached.txt').write_text('\n'.join(uncached))
 print(f'Cache: {len(all_files)-len(uncached)} files hit, {len(uncached)} files need extraction')
 "
@@ -377,6 +384,8 @@ print(f'Merged: {total} nodes, {edges} edges ({len(ast[\"nodes\"])} AST + {len(s
 
 ### Step 4 - Build graph, cluster, analyze, generate outputs
 
+**Before starting:** the code blocks below pass `directed=IS_DIRECTED` to `build_from_json()`. Replace `IS_DIRECTED` with `True` if `--directed` was given (builds a `DiGraph` preserving edge direction source->target), otherwise `False` (the default undirected `Graph`). Substitute it everywhere it appears, the same way you substitute `INPUT_PATH` - do not leave the literal `IS_DIRECTED` in the code.
+
 ```bash
 mkdir -p graphify-out
 $(cat graphify-out/.graphify_python) -c "
@@ -391,7 +400,13 @@ from pathlib import Path
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 detection  = json.loads(Path('.graphify_detect.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
+# Guard BEFORE any write: an empty extraction must not clobber a good graph.json /
+# GRAPH_REPORT.md / analysis sidecar. Check immediately after build (#1392).
+if G.number_of_nodes() == 0:
+    print('ERROR: Graph is empty - extraction produced no nodes.')
+    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
+    raise SystemExit(1)
 communities = cluster(G)
 cohesion = score_all(G, communities)
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -401,9 +416,15 @@ labels = {cid: 'Community ' + str(cid) for cid in communities}
 # Placeholder questions - regenerated with real labels in Step 5
 questions = suggest_questions(G, communities, labels)
 
+# Persist the graph first and only write the report/analysis if it actually
+# persisted - to_json refuses to shrink an existing graph.json (#479), and a
+# report describing a graph we did not write would be a lie (#1392).
+wrote = to_json(G, communities, 'graphify-out/graph.json')
+if not wrote:
+    print('ERROR: refused to shrink graphify-out/graph.json (fewer nodes than the existing graph). Run a full rebuild to be safe.')
+    raise SystemExit(1)
 report = generate(G, communities, cohesion, labels, gods, surprises, detection, tokens, 'INPUT_PATH', suggested_questions=questions)
 Path('graphify-out/GRAPH_REPORT.md').write_text(report)
-to_json(G, communities, 'graphify-out/graph.json')
 
 analysis = {
     'communities': {str(k): v for k, v in communities.items()},
@@ -413,10 +434,6 @@ analysis = {
     'questions': questions,
 }
 Path('.graphify_analysis.json').write_text(json.dumps(analysis, indent=2))
-if G.number_of_nodes() == 0:
-    print('ERROR: Graph is empty - extraction produced no nodes.')
-    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
-    raise SystemExit(1)
 print(f'Graph: {G.number_of_nodes()} nodes, {G.number_of_edges()} edges, {len(communities)} communities')
 "
 ```
@@ -444,7 +461,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 detection  = json.loads(Path('.graphify_detect.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -482,7 +499,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -513,7 +530,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -536,7 +553,7 @@ from graphify.build import build_from_json
 from graphify.export import to_cypher
 from pathlib import Path
 
-G = build_from_json(json.loads(Path('.graphify_extract.json').read_text()))
+G = build_from_json(json.loads(Path('.graphify_extract.json').read_text()), directed=IS_DIRECTED)
 to_cypher(G, 'graphify-out/cypher.txt')
 print('cypher.txt written - import with: cypher-shell < graphify-out/cypher.txt')
 "
@@ -554,7 +571,7 @@ from pathlib import Path
 
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 result = push_to_neo4j(G, uri='NEO4J_URI', user='NEO4J_USER', password='NEO4J_PASSWORD', communities=communities)
@@ -577,7 +594,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -598,7 +615,7 @@ from pathlib import Path
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 to_graphml(G, communities, 'graphify-out/graph.graphml')
@@ -795,7 +812,7 @@ G_existing = json_graph.node_link_graph(existing_data, edges='links')
 
 # Load new extraction
 new_extraction = json.loads(Path('.graphify_extract.json').read_text())
-G_new = build_from_json(new_extraction)
+G_new = build_from_json(new_extraction, directed=IS_DIRECTED)
 
 # Merge: new nodes/edges into existing graph
 G_existing.update(G_new)
@@ -819,7 +836,7 @@ from pathlib import Path
 # Load old graph (before update) from backup written before merge
 old_data = json.loads(Path('.graphify_old.json').read_text()) if Path('.graphify_old.json').exists() else None
 new_extract = json.loads(Path('.graphify_extract.json').read_text())
-G_new = build_from_json(new_extract)
+G_new = build_from_json(new_extract, directed=IS_DIRECTED)
 
 if old_data:
     G_old = json_graph.node_link_graph(old_data, edges='links')

tools/skillgen/expected/graphify__skill-devin.md

@@ -243,12 +243,19 @@ from graphify.cache import check_semantic_cache
 from pathlib import Path
 
 detect = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
-all_files = [f for files in detect['files'].values() for f in files]
+# Only content files go to semantic extraction. Code is already covered
+# structurally by the AST pass; flattening every category here makes the
+# extraction step re-read every source file (#1392).
+all_files = [f for cat in ('document', 'paper', 'image') for f in detect['files'].get(cat, [])]
 
 cached_nodes, cached_edges, cached_hyperedges, uncached = check_semantic_cache(all_files)
 
+# Always (re)write the cache file: write hits, else DELETE any leftover from a
+# prior run so Part C never merges a stale .graphify_cached.json (#1392).
 if cached_nodes or cached_edges or cached_hyperedges:
     Path('graphify-out/.graphify_cached.json').write_text(json.dumps({'nodes': cached_nodes, 'edges': cached_edges, 'hyperedges': cached_hyperedges}))
+else:
+    Path('graphify-out/.graphify_cached.json').unlink(missing_ok=True)
 Path('graphify-out/.graphify_uncached.txt').write_text('\n'.join(uncached))
 print(f'Cache: {len(all_files)-len(uncached)} files hit, {len(uncached)} files need extraction')
 "
@@ -442,6 +449,8 @@ print(f'Merged: {total} nodes, {edges} edges ({len(ast[\"nodes\"])} AST + {len(s
 
 ### Step 4 - Build graph, cluster, analyze, generate outputs
 
+**Before starting:** the code blocks below pass `directed=IS_DIRECTED` to `build_from_json()`. Replace `IS_DIRECTED` with `True` if `--directed` was given (builds a `DiGraph` preserving edge direction source->target), otherwise `False` (the default undirected `Graph`). Substitute it everywhere it appears, the same way you substitute `INPUT_PATH` - do not leave the literal `IS_DIRECTED` in the code.
+
 ```bash
 mkdir -p graphify-out
 $(cat graphify-out/.graphify_python) -c "
@@ -456,7 +465,13 @@ from pathlib import Path
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 detection  = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
+# Guard BEFORE any write: an empty extraction must not clobber a good graph.json /
+# GRAPH_REPORT.md / analysis sidecar. Check immediately after build (#1392).
+if G.number_of_nodes() == 0:
+    print('ERROR: Graph is empty - extraction produced no nodes.')
+    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
+    raise SystemExit(1)
 communities = cluster(G)
 cohesion = score_all(G, communities)
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -466,9 +481,15 @@ labels = {cid: 'Community ' + str(cid) for cid in communities}
 # Placeholder questions - regenerated with real labels in Step 5
 questions = suggest_questions(G, communities, labels)
 
+# Persist the graph first and only write the report/analysis if it actually
+# persisted - to_json refuses to shrink an existing graph.json (#479), and a
+# report describing a graph we did not write would be a lie (#1392).
+wrote = to_json(G, communities, 'graphify-out/graph.json')
+if not wrote:
+    print('ERROR: refused to shrink graphify-out/graph.json (fewer nodes than the existing graph). Run a full rebuild to be safe.')
+    raise SystemExit(1)
 report = generate(G, communities, cohesion, labels, gods, surprises, detection, tokens, 'INPUT_PATH', suggested_questions=questions)
 Path('graphify-out/GRAPH_REPORT.md').write_text(report)
-to_json(G, communities, 'graphify-out/graph.json')
 
 analysis = {
     'communities': {str(k): v for k, v in communities.items()},
@@ -478,10 +499,6 @@ analysis = {
     'questions': questions,
 }
 Path('graphify-out/.graphify_analysis.json').write_text(json.dumps(analysis, indent=2))
-if G.number_of_nodes() == 0:
-    print('ERROR: Graph is empty - extraction produced no nodes.')
-    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
-    raise SystemExit(1)
 print(f'Graph: {G.number_of_nodes()} nodes, {G.number_of_edges()} edges, {len(communities)} communities')
 "
 ```
@@ -509,7 +526,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 detection  = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -547,7 +564,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -578,7 +595,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -632,7 +649,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -655,7 +672,7 @@ from graphify.build import build_from_json
 from graphify.export import to_cypher
 from pathlib import Path
 
-G = build_from_json(json.loads(Path('graphify-out/.graphify_extract.json').read_text()))
+G = build_from_json(json.loads(Path('graphify-out/.graphify_extract.json').read_text()), directed=IS_DIRECTED)
 to_cypher(G, 'graphify-out/cypher.txt')
 print('cypher.txt written - import with: cypher-shell < graphify-out/cypher.txt')
 "
@@ -672,7 +689,7 @@ from pathlib import Path
 
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 result = push_to_neo4j(G, uri='NEO4J_URI', user='NEO4J_USER', password='NEO4J_PASSWORD', communities=communities)
@@ -695,7 +712,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -716,7 +733,7 @@ from pathlib import Path
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 to_graphml(G, communities, 'graphify-out/graph.graphml')
@@ -932,7 +949,7 @@ G_existing = json_graph.node_link_graph(existing_data, edges='links')
 
 # Load new extraction
 new_extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
-G_new = build_from_json(new_extraction)
+G_new = build_from_json(new_extraction, directed=IS_DIRECTED)
 
 # Merge: new nodes/edges into existing graph
 G_existing.update(G_new)
@@ -955,7 +972,7 @@ from pathlib import Path
 
 old_data = json.loads(Path('graphify-out/.graphify_old.json').read_text()) if Path('graphify-out/.graphify_old.json').exists() else None
 new_extract = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
-G_new = build_from_json(new_extract)
+G_new = build_from_json(new_extract, directed=IS_DIRECTED)
 
 if old_data:
     G_old = json_graph.node_link_graph(old_data, edges='links')

tools/skillgen/fragments/core/aider.md

@@ -226,12 +226,19 @@ from graphify.cache import check_semantic_cache
 from pathlib import Path
 
 detect = json.loads(Path('.graphify_detect.json').read_text())
-all_files = [f for files in detect['files'].values() for f in files]
+# Only content files go to semantic extraction. Code is already covered
+# structurally by the AST pass; flattening every category here makes the
+# extraction step re-read every source file (#1392).
+all_files = [f for cat in ('document', 'paper', 'image') for f in detect['files'].get(cat, [])]
 
 cached_nodes, cached_edges, cached_hyperedges, uncached = check_semantic_cache(all_files)
 
+# Always (re)write the cache file: write hits, else DELETE any leftover from a
+# prior run so Part C never merges a stale .graphify_cached.json (#1392).
 if cached_nodes or cached_edges or cached_hyperedges:
     Path('.graphify_cached.json').write_text(json.dumps({'nodes': cached_nodes, 'edges': cached_edges, 'hyperedges': cached_hyperedges}))
+else:
+    Path('.graphify_cached.json').unlink(missing_ok=True)
 Path('.graphify_uncached.txt').write_text('\n'.join(uncached))
 print(f'Cache: {len(all_files)-len(uncached)} files hit, {len(uncached)} files need extraction')
 "
@@ -377,6 +384,8 @@ print(f'Merged: {total} nodes, {edges} edges ({len(ast[\"nodes\"])} AST + {len(s
 
 ### Step 4 - Build graph, cluster, analyze, generate outputs
 
+**Before starting:** the code blocks below pass `directed=IS_DIRECTED` to `build_from_json()`. Replace `IS_DIRECTED` with `True` if `--directed` was given (builds a `DiGraph` preserving edge direction source->target), otherwise `False` (the default undirected `Graph`). Substitute it everywhere it appears, the same way you substitute `INPUT_PATH` - do not leave the literal `IS_DIRECTED` in the code.
+
 ```bash
 mkdir -p graphify-out
 $(cat graphify-out/.graphify_python) -c "
@@ -391,7 +400,13 @@ from pathlib import Path
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 detection  = json.loads(Path('.graphify_detect.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
+# Guard BEFORE any write: an empty extraction must not clobber a good graph.json /
+# GRAPH_REPORT.md / analysis sidecar. Check immediately after build (#1392).
+if G.number_of_nodes() == 0:
+    print('ERROR: Graph is empty - extraction produced no nodes.')
+    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
+    raise SystemExit(1)
 communities = cluster(G)
 cohesion = score_all(G, communities)
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -401,9 +416,15 @@ labels = {cid: 'Community ' + str(cid) for cid in communities}
 # Placeholder questions - regenerated with real labels in Step 5
 questions = suggest_questions(G, communities, labels)
 
+# Persist the graph first and only write the report/analysis if it actually
+# persisted - to_json refuses to shrink an existing graph.json (#479), and a
+# report describing a graph we did not write would be a lie (#1392).
+wrote = to_json(G, communities, 'graphify-out/graph.json')
+if not wrote:
+    print('ERROR: refused to shrink graphify-out/graph.json (fewer nodes than the existing graph). Run a full rebuild to be safe.')
+    raise SystemExit(1)
 report = generate(G, communities, cohesion, labels, gods, surprises, detection, tokens, 'INPUT_PATH', suggested_questions=questions)
 Path('graphify-out/GRAPH_REPORT.md').write_text(report)
-to_json(G, communities, 'graphify-out/graph.json')
 
 analysis = {
     'communities': {str(k): v for k, v in communities.items()},
@@ -413,10 +434,6 @@ analysis = {
     'questions': questions,
 }
 Path('.graphify_analysis.json').write_text(json.dumps(analysis, indent=2))
-if G.number_of_nodes() == 0:
-    print('ERROR: Graph is empty - extraction produced no nodes.')
-    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
-    raise SystemExit(1)
 print(f'Graph: {G.number_of_nodes()} nodes, {G.number_of_edges()} edges, {len(communities)} communities')
 "
 ```
@@ -444,7 +461,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 detection  = json.loads(Path('.graphify_detect.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -482,7 +499,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -513,7 +530,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -536,7 +553,7 @@ from graphify.build import build_from_json
 from graphify.export import to_cypher
 from pathlib import Path
 
-G = build_from_json(json.loads(Path('.graphify_extract.json').read_text()))
+G = build_from_json(json.loads(Path('.graphify_extract.json').read_text()), directed=IS_DIRECTED)
 to_cypher(G, 'graphify-out/cypher.txt')
 print('cypher.txt written - import with: cypher-shell < graphify-out/cypher.txt')
 "
@@ -554,7 +571,7 @@ from pathlib import Path
 
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 result = push_to_neo4j(G, uri='NEO4J_URI', user='NEO4J_USER', password='NEO4J_PASSWORD', communities=communities)
@@ -577,7 +594,7 @@ extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('.graphify_labels.json').read_text()) if Path('.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -598,7 +615,7 @@ from pathlib import Path
 extraction = json.loads(Path('.graphify_extract.json').read_text())
 analysis   = json.loads(Path('.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 to_graphml(G, communities, 'graphify-out/graph.graphml')
@@ -795,7 +812,7 @@ G_existing = json_graph.node_link_graph(existing_data, edges='links')
 
 # Load new extraction
 new_extraction = json.loads(Path('.graphify_extract.json').read_text())
-G_new = build_from_json(new_extraction)
+G_new = build_from_json(new_extraction, directed=IS_DIRECTED)
 
 # Merge: new nodes/edges into existing graph
 G_existing.update(G_new)
@@ -819,7 +836,7 @@ from pathlib import Path
 # Load old graph (before update) from backup written before merge
 old_data = json.loads(Path('.graphify_old.json').read_text()) if Path('.graphify_old.json').exists() else None
 new_extract = json.loads(Path('.graphify_extract.json').read_text())
-G_new = build_from_json(new_extract)
+G_new = build_from_json(new_extract, directed=IS_DIRECTED)
 
 if old_data:
     G_old = json_graph.node_link_graph(old_data, edges='links')

tools/skillgen/fragments/core/devin.md

@@ -243,12 +243,19 @@ from graphify.cache import check_semantic_cache
 from pathlib import Path
 
 detect = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
-all_files = [f for files in detect['files'].values() for f in files]
+# Only content files go to semantic extraction. Code is already covered
+# structurally by the AST pass; flattening every category here makes the
+# extraction step re-read every source file (#1392).
+all_files = [f for cat in ('document', 'paper', 'image') for f in detect['files'].get(cat, [])]
 
 cached_nodes, cached_edges, cached_hyperedges, uncached = check_semantic_cache(all_files)
 
+# Always (re)write the cache file: write hits, else DELETE any leftover from a
+# prior run so Part C never merges a stale .graphify_cached.json (#1392).
 if cached_nodes or cached_edges or cached_hyperedges:
     Path('graphify-out/.graphify_cached.json').write_text(json.dumps({'nodes': cached_nodes, 'edges': cached_edges, 'hyperedges': cached_hyperedges}))
+else:
+    Path('graphify-out/.graphify_cached.json').unlink(missing_ok=True)
 Path('graphify-out/.graphify_uncached.txt').write_text('\n'.join(uncached))
 print(f'Cache: {len(all_files)-len(uncached)} files hit, {len(uncached)} files need extraction')
 "
@@ -442,6 +449,8 @@ print(f'Merged: {total} nodes, {edges} edges ({len(ast[\"nodes\"])} AST + {len(s
 
 ### Step 4 - Build graph, cluster, analyze, generate outputs
 
+**Before starting:** the code blocks below pass `directed=IS_DIRECTED` to `build_from_json()`. Replace `IS_DIRECTED` with `True` if `--directed` was given (builds a `DiGraph` preserving edge direction source->target), otherwise `False` (the default undirected `Graph`). Substitute it everywhere it appears, the same way you substitute `INPUT_PATH` - do not leave the literal `IS_DIRECTED` in the code.
+
 ```bash
 mkdir -p graphify-out
 $(cat graphify-out/.graphify_python) -c "
@@ -456,7 +465,13 @@ from pathlib import Path
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 detection  = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
+# Guard BEFORE any write: an empty extraction must not clobber a good graph.json /
+# GRAPH_REPORT.md / analysis sidecar. Check immediately after build (#1392).
+if G.number_of_nodes() == 0:
+    print('ERROR: Graph is empty - extraction produced no nodes.')
+    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
+    raise SystemExit(1)
 communities = cluster(G)
 cohesion = score_all(G, communities)
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -466,9 +481,15 @@ labels = {cid: 'Community ' + str(cid) for cid in communities}
 # Placeholder questions - regenerated with real labels in Step 5
 questions = suggest_questions(G, communities, labels)
 
+# Persist the graph first and only write the report/analysis if it actually
+# persisted - to_json refuses to shrink an existing graph.json (#479), and a
+# report describing a graph we did not write would be a lie (#1392).
+wrote = to_json(G, communities, 'graphify-out/graph.json')
+if not wrote:
+    print('ERROR: refused to shrink graphify-out/graph.json (fewer nodes than the existing graph). Run a full rebuild to be safe.')
+    raise SystemExit(1)
 report = generate(G, communities, cohesion, labels, gods, surprises, detection, tokens, 'INPUT_PATH', suggested_questions=questions)
 Path('graphify-out/GRAPH_REPORT.md').write_text(report)
-to_json(G, communities, 'graphify-out/graph.json')
 
 analysis = {
     'communities': {str(k): v for k, v in communities.items()},
@@ -478,10 +499,6 @@ analysis = {
     'questions': questions,
 }
 Path('graphify-out/.graphify_analysis.json').write_text(json.dumps(analysis, indent=2))
-if G.number_of_nodes() == 0:
-    print('ERROR: Graph is empty - extraction produced no nodes.')
-    print('Possible causes: all files were skipped, binary-only corpus, or extraction failed.')
-    raise SystemExit(1)
 print(f'Graph: {G.number_of_nodes()} nodes, {G.number_of_edges()} edges, {len(communities)} communities')
 "
 ```
@@ -509,7 +526,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 detection  = json.loads(Path('graphify-out/.graphify_detect.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 tokens = {'input': extraction.get('input_tokens', 0), 'output': extraction.get('output_tokens', 0)}
@@ -547,7 +564,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -578,7 +595,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -632,7 +649,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 cohesion = {int(k): v for k, v in analysis['cohesion'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
@@ -655,7 +672,7 @@ from graphify.build import build_from_json
 from graphify.export import to_cypher
 from pathlib import Path
 
-G = build_from_json(json.loads(Path('graphify-out/.graphify_extract.json').read_text()))
+G = build_from_json(json.loads(Path('graphify-out/.graphify_extract.json').read_text()), directed=IS_DIRECTED)
 to_cypher(G, 'graphify-out/cypher.txt')
 print('cypher.txt written - import with: cypher-shell < graphify-out/cypher.txt')
 "
@@ -672,7 +689,7 @@ from pathlib import Path
 
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 result = push_to_neo4j(G, uri='NEO4J_URI', user='NEO4J_USER', password='NEO4J_PASSWORD', communities=communities)
@@ -695,7 +712,7 @@ extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 labels_raw = json.loads(Path('graphify-out/.graphify_labels.json').read_text()) if Path('graphify-out/.graphify_labels.json').exists() else {}
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 labels = {int(k): v for k, v in labels_raw.items()}
 
@@ -716,7 +733,7 @@ from pathlib import Path
 extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
 analysis   = json.loads(Path('graphify-out/.graphify_analysis.json').read_text())
 
-G = build_from_json(extraction)
+G = build_from_json(extraction, directed=IS_DIRECTED)
 communities = {int(k): v for k, v in analysis['communities'].items()}
 
 to_graphml(G, communities, 'graphify-out/graph.graphml')
@@ -932,7 +949,7 @@ G_existing = json_graph.node_link_graph(existing_data, edges='links')
 
 # Load new extraction
 new_extraction = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
-G_new = build_from_json(new_extraction)
+G_new = build_from_json(new_extraction, directed=IS_DIRECTED)
 
 # Merge: new nodes/edges into existing graph
 G_existing.update(G_new)
@@ -955,7 +972,7 @@ from pathlib import Path
 
 old_data = json.loads(Path('graphify-out/.graphify_old.json').read_text()) if Path('graphify-out/.graphify_old.json').exists() else None
 new_extract = json.loads(Path('graphify-out/.graphify_extract.json').read_text())
-G_new = build_from_json(new_extract)
+G_new = build_from_json(new_extract, directed=IS_DIRECTED)
 
 if old_data:
     G_old = json_graph.node_link_graph(old_data, edges='links')

tools/skillgen/gen.py

@@ -27,6 +27,7 @@ import argparse
 import re
 import subprocess
 import sys
+from collections import Counter
 try:
     import tomllib  # Python 3.11+ stdlib
 except ModuleNotFoundError:  # Python 3.10 - graphify supports >=3.10
@@ -672,8 +673,20 @@ def schema_singleton(platforms: dict[str, Platform]) -> list[str]:
 
 
 def _is_enum_line(line: str) -> bool:
-    """Whether a rendered line carries the unified six-value file_type enum."""
-    return ENUM_VALUES in line or ENUM_PROSE in line
+    """Whether a line carries the file_type enum (its v8 or unified form).
+
+    The unified six-value enum (``ENUM_VALUES``/``ENUM_PROSE``) and the v8
+    five-value form it replaced both match here, so the round-trip's multiset
+    diff classifies the removed-v8 line as well as the added-rendered line. The
+    five-value schema string is a prefix of the six-value one, and both prose
+    forms open with the same ``file_type:"rationale"`` guidance clause.
+    """
+    return (
+        ENUM_VALUES in line
+        or ENUM_PROSE in line
+        or "code|document|paper|image|rationale" in line
+        or 'file_type:"rationale"` for concept-like nodes' in line
+    )
 
 
 def _is_frontmatter_description_line(line: str) -> bool:
@@ -695,8 +708,14 @@ def _is_chunk_cleanup_line(line: str) -> bool:
     ``rm`` and deletes the chunk files with ``find ... -delete`` instead. That
     rewrite touches the single cleanup line in place (no line added or removed),
     so it joins the enum and description unifications as an allowed monolith diff.
+    Both the v8 form (bare ``.graphify_chunk_*.json`` glob, removed) and the fixed
+    form (``find ... -delete``, added) match here so the multiset diff classifies
+    each side of the change.
     """
-    return line.lstrip().startswith("rm -f") and "find " in line and "-name '.graphify_chunk_" in line
+    s = line.lstrip()
+    if not s.startswith("rm -f"):
+        return False
+    return ".graphify_chunk_*.json" in line or ("find " in line and "-name '.graphify_chunk_" in line)
 
 
 def _is_trigger_line(line: str) -> bool:
@@ -708,48 +727,135 @@ def _is_trigger_line(line: str) -> bool:
     return line.strip().startswith("trigger:")
 
 
+def _is_directed_fix_line(line: str) -> bool:
+    """Whether a line is part of the ``--directed`` propagation fix (#1392).
+
+    The monolith runbooks built every graph undirected, so a ``--directed`` run
+    silently collapsed reciprocal A<->B edges. Every ``build_from_json(...)`` call
+    now threads ``directed=IS_DIRECTED`` and a prose line tells the agent to
+    substitute it like ``INPUT_PATH``. Both the old bare call (removed) and the
+    new threaded call (added) match here, plus the substitution instruction.
+    """
+    return (
+        "build_from_json(" in line and "import" not in line
+    ) or "directed=IS_DIRECTED" in line or (
+        "IS_DIRECTED" in line and "Substitute it everywhere" in line
+    )
+
+
+def _is_content_scope_fix_line(line: str) -> bool:
+    """Whether a line is part of the content-only semantic scope fix (#1392).
+
+    Flattening every detect category fed code files (already covered by the AST
+    pass) back to the semantic step. The fix scopes to document/paper/image.
+    """
+    return (
+        "detect['files'].values()" in line
+        or "for cat in ('document', 'paper', 'image')" in line
+        or "Only content files go to semantic extraction" in line
+        or "structurally by the AST pass" in line
+        or "extraction step re-read every source file" in line
+    )
+
+
+def _is_cache_unlink_fix_line(line: str) -> bool:
+    """Whether a line is part of the stale-cache unlink fix (#1392).
+
+    The cache file was written only on a hit, so a miss left a prior run's
+    ``.graphify_cached.json`` for Part C to merge. The miss branch now deletes it.
+    """
+    return (
+        ".graphify_cached.json').unlink(missing_ok=True)" in line
+        or line.strip() == "else:"
+        or "Always (re)write the cache file" in line
+        or "stale .graphify_cached.json" in line
+    )
+
+
+def _is_zero_node_guard_fix_line(line: str) -> bool:
+    """Whether a line is part of the zero-node / shrink-guard ordering fix (#1392).
+
+    Step 4 wrote GRAPH_REPORT.md, graph.json and the analysis sidecar *before*
+    the zero-node guard, so an empty extraction clobbered a good graph; and the
+    report was written even when ``to_json`` refused to shrink (#479). The guard
+    now runs before any write and the report/analysis are gated on ``to_json``.
+    Both the old (removed) and new (added) forms of these lines match here.
+    """
+    s = line.strip()
+    return (
+        "number_of_nodes() == 0" in line
+        or "Graph is empty - extraction produced no nodes" in line
+        or s.startswith("print('Possible causes:")
+        or s == "raise SystemExit(1)"
+        or "to_json(G, communities," in line
+        or s == "if not wrote:"
+        or "refused to shrink graphify-out/graph.json" in line
+        or "Guard BEFORE any write" in line
+        or "GRAPH_REPORT.md / analysis sidecar" in line
+        or "Persist the graph first" in line
+        or "to_json refuses to shrink an existing graph.json" in line
+        or "report describing a graph we did not write" in line
+    )
+
+
+# Every line that may differ between a rendered monolith and its pristine v8
+# baseline. Each predicate documents one sanctioned change-class; a blank line is
+# allowed because the multi-line fix blocks insert spacing. Anything else failing
+# all of these is an unsanctioned drift the round-trip must catch.
+_SANCTIONED_MONOLITH_DIFFS = (
+    _is_enum_line,
+    _is_frontmatter_description_line,
+    _is_chunk_cleanup_line,
+    _is_directed_fix_line,
+    _is_content_scope_fix_line,
+    _is_cache_unlink_fix_line,
+    _is_zero_node_guard_fix_line,
+)
+
+
+def _is_sanctioned_monolith_diff(line: str) -> bool:
+    """Whether a single added/removed monolith line is an allowed change."""
+    return not line.strip() or any(pred(line) for pred in _SANCTIONED_MONOLITH_DIFFS)
+
+
 def monolith_roundtrip(platform: Platform) -> list[str]:
     """Assert a monolith renders diff-clean vs its v8 blob modulo allowed changes.
 
-    Two classes of line are allowed to differ between the rendered monolith and
-    the v8 source: the file_type enum lines (unified to the six-value superset)
-    and the frontmatter ``description`` line (unified across all platforms for
-    discovery). Every other line must match byte for byte.
+    The monolith bodies are hand-maintained single files frozen against a pinned
+    pristine v8 blob (``roundtrip_ref``); this is the guard that stops an
+    arbitrary edit (even a blessed one) from drifting them. Sanctioned changes are
+    enumerated as predicates in ``_SANCTIONED_MONOLITH_DIFFS``: the file_type enum
+    unification, the unified frontmatter description, the chunk-cleanup rewrite
+    (#1172), and the four #1392 runbook fixes (directed propagation, content-only
+    semantic scope, stale-cache unlink, and the zero-node/shrink-guard ordering).
+
+    The comparison is a multiset diff, not a positional zip: a line whose text is
+    unchanged but merely *moved* (the report-write line shifted below ``to_json``
+    in the ordering fix) cancels out and is not flagged. Only lines whose content
+    is genuinely added or removed are checked, and each must be sanctioned.
     """
     if platform.bucket != "monolith":
         return []
     if platform.roundtrip_ref is None:
         return [f"[{platform.key}] monolith is missing roundtrip_ref"]
 
-    rendered = render(platform)[0].content
-    original = _normalise(_git_show(platform.roundtrip_ref))
+    rendered_lines = render(platform)[0].content.splitlines()
+    # Strip trigger lines from the original — they are non-spec and their removal
+    # (#1180) is a permitted diff.
+    original_lines = [
+        l for l in _normalise(_git_show(platform.roundtrip_ref)).splitlines()
+        if not _is_trigger_line(l)
+    ]
 
-    rendered_lines = rendered.splitlines()
-    # Strip trigger lines from the original before comparing — they are non-spec
-    # and their removal (#1180) is a permitted diff. Filter here so the line-count
-    # check and the per-line zip both operate on the same reduced set.
-    original_lines = [l for l in original.splitlines() if not _is_trigger_line(l)]
+    added = Counter(rendered_lines) - Counter(original_lines)
+    removed = Counter(original_lines) - Counter(rendered_lines)
 
     problems: list[str] = []
-    if len(rendered_lines) != len(original_lines):
-        problems.append(
-            f"[{platform.key}] line count differs: rendered {len(rendered_lines)} vs v8 {len(original_lines)} "
-            "(the only allowed changes are the enum line(s), the description line, "
-            "the chunk-cleanup rewrite, and trigger: removal — none must add or remove other lines)"
-        )
-        return problems
-
-    for i, (r, o) in enumerate(zip(rendered_lines, original_lines), start=1):
-        if r == o:
-            continue
-        # The permitted diffs are the enum unification, the unified description,
-        # the shell-agnostic chunk-cleanup rewrite (#1172), and trigger removal (#1180).
-        if _is_enum_line(r) or _is_frontmatter_description_line(r) or _is_chunk_cleanup_line(r):
+    for line in list(added.elements()) + list(removed.elements()):
+        if _is_sanctioned_monolith_diff(line):
             continue
         problems.append(
-            f"[{platform.key}] line {i} differs and is not an enum or description unification:\n"
-            f"    v8:       {o!r}\n"
-            f"    rendered: {r!r}"
+            f"[{platform.key}] unsanctioned monolith change vs pristine v8: {line!r}"
         )
     return problems