🛡️ Sentinel: [CRITICAL] Fix data leak in public export endpoints

.Jules/sentinel.md

@@ -1,4 +1,9 @@
-## $(date +%Y-%m-%d) - Redact Sensitive Schema Comments in Public Shares
+## 2026-07-14 - Redact Sensitive Schema Comments in Public Shares
 **Vulnerability:** Publicly shared schema snapshots (via `/api/share/...`) returned the entire JSON payload, which could expose sensitive internal schema comments (`comment`, `relation_comment`, `column_comment`) or sensitive data in `example_value` fields.
 **Learning:** When generating share links, only specific fields should be exposed, but we export the entire `snapshot_json` from the database. A recursive sanitizer function must be applied to scrub IDOR/data leakage vectors before returning the JSON payload.
 **Prevention:** Apply a recursive masking function (`_redact_sensitive_snapshot_fields`) on database JSON artifacts in read-only public endpoints.
+
+## 2026-07-14 - [CRITICAL] Fix Data Leak in Public Export Endpoints
+**Vulnerability:** Unauthenticated share link endpoints for SQL export, DB reversing spec, and index design spec were directly passing un-redacted snapshot JSON data to the generators. This could expose sensitive metadata like `comment`, `relation_comment`, `column_comment`, and `example_value`.
+**Learning:** Even when exposing data through format transformers (like SQL generators or LLM prompts), any sensitive metadata fields included in the base JSON payload can leak if not explicitly redacted before transformation.
+**Prevention:** Always apply the `_redact_sensitive_snapshot_fields` utility (or equivalent redaction logic) to JSON payloads before passing them to any export or transformation function exposed via unauthenticated endpoints.

.gitignore

@@ -31,3 +31,5 @@ frontend/coverage/
 registered_agents.json
 task_agent_mapping.json
 backend/venv/
+myenv/
+test_venv/

backend/app/api/share.py

@@ -2,6 +2,7 @@
 
 import datetime as dt
 import uuid
+from typing import cast
 
 from fastapi import APIRouter, Depends, HTTPException, Query
 from fastapi.responses import PlainTextResponse
@@ -29,19 +30,39 @@
 router = APIRouter(prefix="/api", tags=["share"])
 
 
+import re
+from app.dsn_redaction import redact_dsn_error_message
+
+import re
+from app.dsn_redaction import redact_dsn_error_message
+
 def _redact_sensitive_snapshot_fields(
     data: dict | list | str | int | float | bool | None,
 ) -> dict | list | str | int | float | bool | None:
     """Redact sensitive fields from snapshot JSON payload when shared publicly."""
     if isinstance(data, dict):
         return {
             k: "***"
-            if k in {"comment", "relation_comment", "column_comment", "example_value"}
+            if isinstance(k, str) and (
+                k.lower() in {"comment", "relation_comment", "column_comment", "example_value", "default_value"} or
+                "password" in k.lower() or
+                "secret" in k.lower() or
+                "credential" in k.lower() or
+                "token" in k.lower() or
+                "key" in k.lower()
+            )
             else _redact_sensitive_snapshot_fields(v)
             for k, v in data.items()
         }
     elif isinstance(data, list):
         return [_redact_sensitive_snapshot_fields(v) for v in data]
+    elif isinstance(data, str):
+        if "://" in data or "@" in data:
+            try:
+                from app.dsn_redaction import redact_dsn_error_message
+                return redact_dsn_error_message(data, "")
+            except Exception:
+                pass
     return data
 
 
@@ -140,7 +161,7 @@ async def get_shared_snapshot(
         "status": snap.status,
         "schema_filter": snap.schema_filter,
         "error_message": snap.error_message,
-        "snapshot_json": _redact_sensitive_snapshot_fields(data.snapshot_json)
+        "snapshot_json": cast(dict, _redact_sensitive_snapshot_fields(data.snapshot_json))
         if data
         else None,
     }
@@ -172,7 +193,7 @@ async def export_shared_snapshot_sql(
     data = await session.get(SchemaSnapshotData, schema_snapshot_uuid)
     if data is None:
         return "-- snapshot data not found\n"
-    return snapshot_json_to_sql(data.snapshot_json, target_dialect=dialect)
+    return snapshot_json_to_sql(cast(dict, _redact_sensitive_snapshot_fields(data.snapshot_json)), target_dialect=dialect)
 
 
 @router.get(
@@ -203,7 +224,7 @@ async def export_shared_snapshot_reversing_spec(
         return "# DB Reversing Specification\n\nSnapshot data not found.\n"
     if mode == "llm-draft":
         try:
-            return await generate_reversing_llm_draft(data.snapshot_json)
+            return await generate_reversing_llm_draft(cast(dict, _redact_sensitive_snapshot_fields(data.snapshot_json)))
         except LlmConfigurationError as exc:
             raise HTTPException(
                 status_code=503, detail="LLM configuration error"
@@ -212,7 +233,7 @@ async def export_shared_snapshot_reversing_spec(
             raise HTTPException(
                 status_code=502, detail="LLM provider request failed"
             ) from exc
-    return generate_reversing_spec(data.snapshot_json, mode=mode)
+    return generate_reversing_spec(cast(dict, _redact_sensitive_snapshot_fields(data.snapshot_json)), mode=mode)
 
 
 @router.get(
@@ -243,7 +264,7 @@ async def export_shared_snapshot_index_design(
         return "# ERD Index Design\n\nSnapshot data not found.\n"
     if mode == "llm-draft":
         try:
-            return await generate_index_design_llm_draft(data.snapshot_json)
+            return await generate_index_design_llm_draft(cast(dict, _redact_sensitive_snapshot_fields(data.snapshot_json)))
         except LlmConfigurationError as exc:
             raise HTTPException(
                 status_code=503, detail="LLM configuration error"
@@ -252,4 +273,4 @@ async def export_shared_snapshot_index_design(
             raise HTTPException(
                 status_code=502, detail="LLM provider request failed"
             ) from exc
-    return generate_index_design_spec(data.snapshot_json, mode=mode)
+    return generate_index_design_spec(cast(dict, _redact_sensitive_snapshot_fields(data.snapshot_json)), mode=mode)