Fix official policy selection and persistent attachment gaps

Read the OFAC action body and use focused macro policy phrases, including national budgets, appropriations and reserve policies. Keep unknown image text as an explicit gap instead of treating it as complete or irrelevant.\n\nPreserve meeting attachment retries and derive pending/exhausted document counts from saved source cursors. Cover real selection negatives, positive macro boundaries and attachment failures with isolated regressions.

deploy/README.md

@@ -276,6 +276,22 @@ attachment text are searchable and versioned. Date-only releases are displayed a
 dates, not precise release instants. Backfilled documents describe the version
 available at capture time; they are not historical point-in-time evidence.
 
+Selection rule `us-policy-v2` uses explicit national budget, continuing/consolidated
+appropriations, discount-rate and reserve-policy phrases instead of standalone
+`budget`, `fiscal` or `quantitative`. Federal Register macro screening uses the title
+to avoid incidental references in unrelated regulations; crypto screening still
+reads the complete body and extracted attachments.
+
+Image attachments remain an explicit text-extraction gap, including when their
+format is PNG, JPEG or GIF. They do not establish completeness or irrelevance.
+Relevant documents remain `partial`; documents whose relevance is still unknown
+are kept in bounded attachment retries rather than inserted as unrelated news.
+Congress meeting attachment failures retain the same retry evidence as articles.
+The source API derives `attachment_gaps`, `attachment_retry_pending` and
+`attachment_retry_exhausted` from the persisted cursor, independently of the latest
+round's statistics. The console keeps these document counts visible across rounds;
+success clears the gap, while three failed attempts remain visible for investigation.
+
 Validation without production credentials:
 
 - `cd backend && uv run pytest && uvx ruff check --select F,E9,B src tests`