scripts/metadao-scrape.py

@@ -57,6 +57,11 @@ def slugify(text: str, max_len: int = 60) -> str:
     return s.strip("-")[:max_len].rstrip("-")
 
 
+def _yaml_str(s: str) -> str:
+    """Quote-safe YAML string. JSON strings are valid YAML strings."""
+    return json.dumps(s, ensure_ascii=False)
+
+
 def existing_basenames(*dirs: Path) -> set[str]:
     """Collect all .md basenames (without extension) across the given dirs (recursive)."""
     seen: set[str] = set()
@@ -69,7 +74,7 @@ def existing_basenames(*dirs: Path) -> set[str]:
 
 
 PROP_ADDR_RE = re.compile(r"proposal_address:\s*[\"']?([A-Za-z0-9]{32,44})[\"']?")
-URL_ADDR_RE = re.compile(r"(?:futard\.io|metadao\.fi)/[^/\s\"']*proposal/([A-Za-z0-9]{32,44})")
+URL_ADDR_RE = re.compile(r"(?:futard\.io|metadao\.fi)(?:/[^/\s\"']*)*?/proposal/([A-Za-z0-9]{32,44})")
 
 
 def existing_proposal_addresses(*dirs: Path) -> set[str]:
@@ -306,28 +311,31 @@ def build_source_markdown(project: dict, proposal: dict, today: str) -> str:
     body_text = (proposal.get("body_text") or "").strip()
     decoded = proposal.get("decoded") or {}
 
-    # Build YAML frontmatter
+    # Build YAML frontmatter — all free-text values escaped via _yaml_str (json.dumps).
+    # project_slug is constrained to [a-z0-9-] by slugify upstream, but pass through
+    # the same path for consistency.
+    full_title = f"MetaDAO: {project['name']} — {title}"
     fm_lines = [
         "---",
         "type: source",
-        f'title: "MetaDAO: {project["name"]} — {title}"',
-        'author: "metadao.fi"',
-        f'url: "{proposal["url"]}"',
+        f"title: {_yaml_str(full_title)}",
+        f"author: {_yaml_str('metadao.fi')}",
+        f"url: {_yaml_str(proposal['url'])}",
         f"date: {today}",
         "domain: internet-finance",
         "format: data",
         "status: unprocessed",
-        f'tags: [futardio, metadao, futarchy, solana, governance, {project["slug"]}]',
+        f"tags: [futardio, metadao, futarchy, solana, governance, {project['slug']}]",
         "event_type: proposal",
-        f'project_slug: "{project["slug"]}"',
-        f'proposal_address: "{proposal["address"]}"',
+        f"project_slug: {_yaml_str(project['slug'])}",
+        f"proposal_address: {_yaml_str(proposal['address'])}",
     ]
     if proposal.get("status"):
-        fm_lines.append(f'proposal_status: "{proposal["status"]}"')
+        fm_lines.append(f"proposal_status: {_yaml_str(proposal['status'])}")
     if decoded.get("squadsProposal"):
-        fm_lines.append(f'squads_proposal: "{decoded["squadsProposal"]}"')
+        fm_lines.append(f"squads_proposal: {_yaml_str(decoded['squadsProposal'])}")
     if decoded.get("squadsStatus"):
-        fm_lines.append(f'squads_status: "{decoded["squadsStatus"]}"')
+        fm_lines.append(f"squads_status: {_yaml_str(decoded['squadsStatus'])}")
     fm_lines.append("---")
     fm_lines.append("")
 
@@ -458,7 +466,7 @@ def main() -> int:
                 if args.limit and len(written) >= args.limit:
                     log.info("hit limit=%d, stopping", args.limit)
                     browser.close()
-                    print(json.dumps({"written": written, "skipped_existing": skipped_existing}))
+                    print(json.dumps({"written": written, "skipped_existing": skipped_existing, "dry_run": args.dry_run}))
                     return 0
 
         browser.close()