This release adds 3 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
+14 more
ReleasePort's take
Light signalRelease v0.9.3 adds test fixtures, documentation on evidence modes, a product rubric, and consolidates stack‑overview tables.
Why it matters: Test coverage rises to 142 cases; new docs clarify schema values and evaluation criteria, while duplicate README content is eliminated for clearer maintenance.
Summary
AI summaryAdded evidence-mode discipline fixtures, documentation on schema values and a product rubric.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Medium |
Adds 5 golden and 5 failure fixtures to tests/fixtures/evidence_mode/, increasing test count from 131 to 142. Adds 5 golden and 5 failure fixtures to tests/fixtures/evidence_mode/, increasing test count from 131 to 142. Source: llm_adapter@2026-05-22 Confidence: high |
— |
| Feature | Medium |
Adds canonical `docs/evidence-modes.md` describing evidence_mode values, check_evidence_mode_discipline rule, and fixture table. Adds canonical `docs/evidence-modes.md` describing evidence_mode values, check_evidence_mode_discipline rule, and fixture table. Source: llm_adapter@2026-05-22 Confidence: low |
— |
| Feature | Medium |
Adds new `docs/rubric.md` with a 10‑dimension product‑shell rubric and a 6‑point human review checklist. Adds new `docs/rubric.md` with a 10‑dimension product‑shell rubric and a 6‑point human review checklist. Source: llm_adapter@2026-05-22 Confidence: low |
— |
| Refactor | Medium |
Centralizes canonical "Where this fits in the Agenda Intelligence stack" 4‑row table into README.md; removes duplicates from three skill repos. Centralizes canonical "Where this fits in the Agenda Intelligence stack" 4‑row table into README.md; removes duplicates from three skill repos. Source: llm_adapter@2026-05-22 Confidence: high |
— |
Full changelog
Closes the v0.9 release gate — all 9/9 acceptance criteria ✅.
What's in this release
Phase 4b — evidence-mode discipline extended to 5+5 (#43)
tests/fixtures/evidence_mode/now ships 5 golden + 5 failure fixtures (was 3+3), covering sanctions designations, vessel SDN attribution, and EU regulatory effective-date claims as the three determinative-claim domains where the discipline rule matters most. Fixtures generated deterministically bybuild.pyfrom one baseline memo. Test count: 131 → 142.- New
docs/evidence-modes.md— canonical reference for the three schemaevidence_modevalues, the post-hoccheck_evidence_mode_disciplinerule, and the full fixture table. - New
docs/rubric.md— 10-dimension product-shell rubric + 6-point human review checklist (complements the deterministic heuristic scorer inevals/rubric.md).
Phase 8 — 4-layer map deduplicated across the portfolio (#44)
The "Where this fits in the Agenda Intelligence stack" 4-row table is now canonical in this repo's README.md only. The three skill repos drop their local duplicates and link to the canonical anchor:
- global-think-tank-analyst#10
- central-asia-caspian-hybrid-intelligence-skill#12
- gulf-middle-east-hybrid-intelligence-skill#11
Compatibility
No schema, CLI, or MCP tool changes. No new dependencies. Existing requests and memos remain valid.
Install
pip install agenda-intelligence-md==0.9.3
Or pinned wheel from this release:
pip install https://github.com/vassiliylakhonin/agenda-intelligence-md/releases/download/v0.9.3/agenda_intelligence_md-0.9.3-py3-none-any.whl
See CHANGELOG.md for full notes.
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Agenda Intel MD
All releases →Related context
Related tools
Earlier breaking changes
- v0.8.0 MCP tool count increased from 11 to 16, adding five new tools.
Beta — feedback welcome: [email protected]