v1.1.1 - AnalysisBank and confidential workflow hardening

v1.1.1 is a quality-gate patch release for the evidence-readiness runtime. It
does not add a new worker, buyer-facing surface, factual verification, legal or
compliance advice, sanctions screening, or autonomous decision logic.

What changed

• Hardened AnalysisBank from loose reasoning-memory notes into a guarded
  runtime path:
  • lifecycle metadata linting via memory-lint;
  • packaged memory parity under src/agenda_intelligence/data/;
  • deterministic memory-search ranking;
  • retrieval regression fixtures via memory-search-bench;
  • applicability regression fixtures via memory-applicability-bench;
  • bounded REASONING MEMORY selection in analyze;
  • audit.reasoning_memory trace in memos.
• Added memo-quality regression coverage:
  • fixture manifest schema;
  • 5 golden memo fixtures and 8 schema-valid failure fixtures;
  • confidential project-room alias-leak guard;
  • CI smoke for memo-quality-bench.
• Added confidential weekly/status regression coverage:
  • new weekly-delta-bench CLI;
  • weekly-delta fixture manifest schema;
  • one alias-only committee-escalation golden fixture;
  • one named-project leak failure fixture;
  • CI smoke for weekly-delta-bench.
• Documented the AnalysisBank hardening checkpoint:
  • before/after state;
  • runtime path;
  • operator checklist;
  • quality gates;
  • stop rule for further memory mechanics.

What did not change

• The project remains an evidence-readiness and trust-routing layer.
• No factual truth verification was added.
• No legal, compliance, sanctions, financial, investment, procurement, tax, or
  customs advice was added.
• No Cloudflare Worker deployment is part of this release.
• No commercial traction claim is made.

Operator gate

Before editing AnalysisBank cards:

agenda-intelligence memory-lint
agenda-intelligence memory-search-bench tests/fixtures/analysis_bank_retrieval/manifest.json
agenda-intelligence memory-applicability-bench tests/fixtures/analysis_bank_applicability/manifest.json

Before accepting memo or confidential workflow changes:

agenda-intelligence memo-quality-bench tests/fixtures/memo_quality --format json
agenda-intelligence weekly-delta-bench tests/fixtures/weekly_delta --format json
make ci PYTHON=python3

Release decision

The repository is ready for a v1.1.1 tag only if PyPI publication is intended
or the tag-triggered release workflow is changed first. The current
.github/workflows/release.yml publishes to PyPI on any v* tag push.