Docs: Scaling and Strategic Refinement plan for 20-agent pool and Strategy Council

docs/plans/Scaling_and_Strategic_Refinement.md

@@ -0,0 +1,39 @@
+# Plan: Scaling and Strategic Refinement (2026-03-03)
+
+## 1. Goal: Council & Agent Scaling
+Transition CHORUS from a proof-of-concept trio to a robust, 20-agent ecosystem capable of high-level strategic planning and execution.
+
+### Objectives:
+- **Default Council Configuration**: Define a "Strategy Council" consisting of 10+ roles:
+    - Senior Software Architect (Lead)
+    - Systems Analyst
+    - TPM
+    - Security Architect
+    - Developer (x3)
+    - Reviewer
+    - QA Test Engineer
+    - SRE Observability Lead
+- **Generic Agent Pool**: Provision 20 `chrs-agent` instances that can dynamically assume any role from the `chrs-prompts` library based on council needs.
+- **Project Brief Ingestion**: Have the council load and analyze all project briefs in `docs/distos/councils/` to determine the global implementation strategy.
+
+## 2. Goal: Testing & Quality Assurance
+Ensure that the "Agent Inception" loop produces code and documentation that meets human engineering standards.
+
+### Objectives:
+- **Testing Harnesses**: Build specialized crates or scripts to monitor the "Delegation Chain."
+    - Verify that the Architect's high-level intent is correctly partitioned by the TPM.
+    - Verify that sub-tasks contain enough context for the containerized sub-agent to succeed.
+- **Output Audit**: Critically analyze the `stdout` and `execution_results` in Dolt.
+    - If code fails to compile or analyze correctly, identify the missing "Prime Directives."
+    - Refine `chrs-prompts` to include better "Chain of Thought" instructions.
+
+## 3. Goal: Data Analysis & Module Refinement
+Use the scaled-up cluster to stress-test and improve our coordination logic.
+
+### Objectives:
+- **Inter-Agent Comms Analysis**: Analyze the Gossipsub and Mailbox traffic to identify bottlenecks or "Communication Storms."
+- **SLURP Refinement**: Enhance the Context Intelligence layer to better handle "Epistemic Fences" (ensuring agents don't see what they don't need to know).
+- **BUBBLE Refinement**: Improve the Provenance DAG to support "Ancestry Queries"—answering *exactly* why a line of code exists by tracing it back through the sub-agent, the developer, and the original council decision.
+
+---
+*Created by CHORUS CLI Coordinator - March 3, 2026*