v0.8.0 — Agenda Intelligence product shell

v0.8.0 turns this repository into the entry point for the Agenda Intelligence product. Five new MCP tools wrap the existing validation surface with geography routing, system-prompt assembly from bundled reasoning method and regional references, optional direct LLM invocation, and a vendored snapshot of the Global Think Tank Analyst signal archive. The validation layer remains unchanged.

Added — Product-shell MCP tools

• analyze — full pipeline: validates the request against agenda-request.schema.json, routes geography to the relevant in-repo regional / sector references (Central Asia + Caspian, Gulf + Middle East, sanctions), assembles a system prompt from the bundled SKILL.md and reference files, and — when ANTHROPIC_API_KEY is set and the optional anthropic SDK is installed — calls the Anthropic API and validates the returned memo against agenda-memo.schema.json. Without an API key, returns a skeleton memo plus the assembled system_prompt so a host model can complete the analysis.
• validate_memo — schema check against agenda-memo.schema.json for memos produced outside the analyze pipeline.
• list_signals / get_signal — read-only access to a vendored snapshot of the Global Think Tank Analyst signal archive under data/signals/.
• deep_dive — reserved for v2; returns a planned-status message directing callers to analyze with depth: scenario or red_team.

MCP tool count: 11 → 16.

Added — Product request and memo schemas

• schemas/agenda-request.schema.json — input contract: question, decision_context, audience, geography, time_horizon, evidence_mode, depth, output_format. Draft 2020-12, strict.
• schemas/agenda-memo.schema.json — output contract: meta (incl. modules_used, gtta_version, timestamp), risk_summary (short / detailed), decision_frame, analysis (facts / assessments / assumptions / unknowns), scenarios with probability ranges, options, recommended_actions, failure_modes, watch_next, audit (validation_score, provenance). Both schemas include validated examples.
• Schemas are exposed in agent-manifest.json under schemas (as agenda_request and agenda_memo) and via a new product block declaring the request/response contract, the five MCP tools, and the explicit live_source_retrieval: false flag.

Added — Vendored signal archive

• scripts/sync_signals.py copies the Global Think Tank Analyst signals/ directory into src/agenda_intelligence/data/signals/, mirror of the schemas/ ↔ data/schemas/ dual-copy pattern.
• tests/test_signal_sync.py enforces parity when a local GTTA checkout is present and is skipped on CI without it.
• Initial snapshot: 10 individual signals plus index.json, feed.json, latest.md, TEMPLATE.md, README.md.

Added — Optional LLM dependency

• pip install agenda-intelligence-md[llm] installs anthropic>=0.40 so analyze can call the Anthropic API directly via ANTHROPIC_API_KEY. No new hard dependency.

Added — Product-shell tests

• tests/test_product_shell.py — 10 integration tests covering geography routing (Kazakhstan → CA-Caspian, global → GTTA only, mixed Iran/UAE → Gulf+ME), validate_memo happy and negative paths, the LLM-invocation branch with mocked Anthropic responses (JSON and non-JSON), and signals + deep_dive coverage. Tests run without network or credentials.

Changed — Repository positioning

• README.md leads with the product framing: Agenda Intelligence is a trusted geopolitical intelligence layer for agentic workflows. Auditable, structured, decision-grade risk memos with evidence discipline that generic LLMs lack. One MCP server. Structured input/output. Built-in validation. The "What this is" section now opens with the MCP product shell. The validator's hardcoded README tagline token has been updated to match.
• MCP.md documents the five new tools with request/response examples and tool-count update (11 → 16).

Unchanged

• Live source retrieval is not implemented. evidence_mode is restricted to reasoning_only, user_provided, and mixed.
• The existing 11 validation tools (validate_brief, validate_evidence, audit_claims, get_protocol, list_lenses, get_lens, source_plan, list_source_categories, source_coverage, score_output, verify_quotes) keep their contracts.
• No legal, compliance, financial, or security advice. No autonomous decision-making. No production-grade guarantees.

Companion repositories

• Global Think Tank Analyst — reasoning method loaded by analyze.
• Central Asia + Caspian Hybrid Intelligence Skill — regional specialist activated when geography matches.
• Gulf + Middle East Hybrid Intelligence Skill — regional specialist activated when geography matches.

Each companion README now names Agenda Intelligence and documents whether the skill activates automatically (regional specialists) or sits behind the analyze tool as the reasoning method (GTTA).

Upgrade notes

Existing CLI users see no behavioral changes. Existing MCP clients see five additional tools in tools/list. The wheel installs the new product layer without requiring the Anthropic SDK; install the [llm] extra and export ANTHROPIC_API_KEY to enable direct API calls from analyze.