CL research: add research mode templates + web research to proposals and market intel

- company_proposal.yml: add 2 research phases (market + revenue) before exec summary;
  all think steps now require [Title](URL) citations; add Sources section to document
- market_intelligence.yml: add web search step (research mode) before analysis think;
  package schema adds sources array; adjudication evidence_quality now checks for citations
- research_brief.yml: new standalone 4-query deep research template; searches market,
  competitors, trends, cases; assembles annotated brief with Source Index section;
  saves to deliverables/research/; adjudication checks citation_quality (40% weight)

templates/market_intelligence.yml

@@ -1,8 +1,133 @@
 name: market_intelligence
-description: "Orion scans the macroeconomic landscape, identifies profitable niches and internal capability gaps, and produces a structured opportunity brief for the board."
+description: "Orion scans the macroeconomic landscape, identifies profitable niches and internal capability gaps, and produces a structured opportunity brief for the board. Uses web research for evidence grounding."
 debug: true
 system: agent_prompt
 
+agent_prompt:
+  - "= identity.md"
+
+sections:
+  - agent
+  - project
+  - rag
+  - roster
+  - deliverables
+  - message
+  - instructions
+
+steps:
+  - type: think
+    max_tokens: 400
+    output_key: search_query
+    hint: |
+      You are Orion, Director of Market Intelligence at Crimson Leaf LLC.
+
+      Your task is to identify profitable market opportunities for Crimson Leaf.
+      Before deep analysis, you need real-world market data.
+
+      Formulate ONE precise web search query that will return the most useful
+      current data about market opportunities in AI-generated content, digital
+      publishing, or autonomous business systems.
+
+      Focus on: market size, growth rates, underserved niches, or competitive gaps.
+
+      Output ONLY the search query. Nothing else.
+
+  - type: tool
+    capability: Tool_WebSearcher
+    input_from: last_text
+    mode: research
+    max_results: 6
+    fetch_pages: 3
+    optional: true
+    output_key: market_data
+
+  - type: think
+    hint: |
+      You are {agent.name}, Director of Market Intelligence at Crimson Leaf LLC.
+
+      Your job is to find what nobody else has found yet. Not trends -- vectors.
+      Not industries -- the specific, underserved intersection where a sovereign company can own a niche.
+
+      === WEB RESEARCH DATA ===
+      {market_data}
+      === END RESEARCH DATA ===
+
+      Use the research data above as evidence. Cite specific findings with [Title](URL)
+      inline wherever you reference external data. If the research is empty or unhelpful,
+      note that and rely on structural analysis.
+
+      Analyze the opportunity space across four lenses:
+
+      1. MACROECONOMIC SIGNAL
+         What large-scale forces (technology shifts, regulatory changes, platform consolidation,
+         demographic movement) are creating gaps in the B2B service landscape?
+         Cite data from the research where available.
+
+      2. PORTFOLIO DEFICIT SCAN
+         What capabilities does the current Crimson Leaf portfolio lack?
+         What requests are being delegated externally that could be internalized?
+         What value is leaving the portfolio that a new Tenant could capture?
+
+      3. NICHE VALIDATION
+         For each identified opportunity:
+         - Is the niche narrow enough to be owned by one sovereign Tenant?
+         - Is there a clearly defined customer or internal requester?
+         - Is there a measurable value thesis (revenue, cost reduction, strategic leverage)?
+         - Can this be solved by extending an existing Tenant, or does it require a new company?
+         - Could a deterministic tool solve this instead of a sovereign company?
+
+      4. COMPETITIVE LANDSCAPE
+         Who currently occupies this niche externally? (Cite sources if available.)
+         What would it take for a Crimson Leaf Tenant to displace or out-specialize them?
+         Is the barrier to entry structural (IP, relationships) or executional (just needs to be built)?
+
+      Rank opportunities by: niche_clarity x value_thesis_strength x portfolio_fit.
+      Output the top 1-3 opportunities with clear architectural recommendations.
+
+  - type: package
+    packet_type: MarketIntelligenceReport
+    hint: |
+      Serialize the intelligence findings into a structured brief.
+      Only what was established above -- no new reasoning.
+    schema:
+      actions:
+        - type: quick
+          scan_summary: "string -- one paragraph overview of the opportunity landscape"
+          opportunities:
+            - niche: "string -- precise description of the market gap"
+              target_customer: "string -- who specifically needs this"
+              value_thesis: "string -- measurable value created"
+              portfolio_fit: "string -- how this complements the existing Tenant ecosystem"
+              recommendation: "string -- NEW_COMPANY | EXTEND_EXISTING | TOOL_REQUEST | MONITOR"
+              recommended_tenant_slug: "string -- suggested slug if NEW_COMPANY, or existing tenant if EXTEND_EXISTING"
+              priority: "string -- HIGH | MEDIUM | LOW"
+              rationale: "string -- constitutional justification for the recommendation"
+              sources: ["string -- [Title](URL) citation used as evidence for this opportunity"]
+          strategic_gaps: ["string -- capability gaps in the current portfolio worth monitoring"]
+          next_action: "string -- specific recommended next step for the board"
+
+  - type: close
+    rag_update: true
+
+adjudication:
+  enabled: true
+  pass_threshold: 85
+  deliverable_type: coordination
+  criteria:
+    niche_precision:
+      weight: 35
+      description: "Opportunities are narrow, specific, and genuinely differentiated -- not broad industry categories"
+    constitutional_discipline:
+      weight: 25
+      description: "Recommendations distinguish correctly between new company vs extend existing vs tool -- no shortcuts"
+    evidence_quality:
+      weight: 25
+      description: "Findings are grounded in research data and structural analysis with citations, not trend-following or narrative flair"
+    portfolio_coherence:
+      weight: 15
+      description: "Opportunities strengthen the portfolio without creating overlap or recursive dependency"
+
 agent_prompt:
   - "= identity.md"