scripts: use capnp store instead of reading markdown directly

Add store_helpers.py with shared helpers that call poc-memory commands
(list-keys, render, journal-tail) instead of globbing ~/.claude/memory/*.md
and parsing section headers.

All 9 Python scripts updated: get_semantic_keys(), get_topic_file_index(),
get_recent_journal(), parse_journal_entries(), read_journal_range(),
collect_topic_stems(), and file preview rendering now go through the store.

This completes the clean switch — no script reads archived markdown files.

scripts/consolidation-loop.py

@@ -81,17 +81,9 @@ def get_health() -> dict:
 
 
 def get_topic_file_index() -> dict[str, list[str]]:
-    """Build index of topic files and their section headers."""
-    index = {}
-    for md in sorted(MEMORY_DIR.glob("*.md")):
-        name = md.name
-        headers = []
-        for line in md.read_text().split('\n'):
-            if line.startswith('## '):
-                slug = re.sub(r'[^a-z0-9-]', '', line[3:].lower().replace(' ', '-'))
-                headers.append(slug)
-        index[name] = headers
-    return index
+    """Build index of topic files and their section headers from the store."""
+    from store_helpers import get_topic_file_index as _get_index
+    return _get_index()
 
 
 def get_graph_structure() -> str:
@@ -123,12 +115,14 @@ def build_crosslink_prompt(round_num: int) -> str:
     graph = get_graph_structure()
     status = get_status()
 
-    # Read a sample of files for context
+    # Read a sample of files from the store
+    from store_helpers import render as _render
     file_previews = ""
-    for f in sorted(MEMORY_DIR.glob("*.md"))[:30]:
-        content = f.read_text()
-        preview = '\n'.join(content.split('\n')[:8])[:400]
-        file_previews += f"\n--- {f.name} ---\n{preview}\n"
+    for fname in sorted(index.keys())[:30]:
+        content = _render(fname)
+        if content:
+            preview = '\n'.join(content.split('\n')[:8])[:400]
+            file_previews += f"\n--- {fname} ---\n{preview}\n"
 
     return f"""You are a cross-link discovery agent (round {round_num}).
 
@@ -210,13 +204,13 @@ Output ONLY the JSON array."""
 
 def build_newfile_prompt(round_num: int) -> str:
     """Build prompt for connecting the new split files."""
-    # Read the new reflection files
+    # Read the new reflection files from the store
+    from store_helpers import render as _render
     new_files = {}
     for name in ['reflections-reading.md', 'reflections-dreams.md', 'reflections-zoom.md',
                  'verus-proofs.md']:
-        path = MEMORY_DIR / name
-        if path.exists():
-            content = path.read_text()
+        content = _render(name)
+        if content:
             new_files[name] = content[:2000]
 
     # Read existing files they should connect to
@@ -224,9 +218,8 @@ def build_newfile_prompt(round_num: int) -> str:
     for name in ['differentiation.md', 'cognitive-modes.md', 'language-theory.md',
                  'discoveries.md', 'inner-life.md', 'design-context-window.md',
                  'design-consolidate.md', 'experiments-on-self.md']:
-        path = MEMORY_DIR / name
-        if path.exists():
-            content = path.read_text()
+        content = _render(name)
+        if content:
             target_files[name] = content[:1500]
 
     graph = get_graph_structure()