Final updates to global rules for the AI agent factory

use-cases/agent-factory-with-subagents/.claude/settings.local.json

@@ -20,7 +20,8 @@
       "Bash(gh issue view:*)",
       "mcp__archon",
       "WebSearch",
-      "Bash(cp:*)"
+      "Bash(cp:*)",
+      "Bash(sed:*)"
     ],
     "deny": []
   }

use-cases/agent-factory-with-subagents/.gitignore

@@ -7,4 +7,5 @@ asana_agent
 asana_manager
 brave_search_agent
 pgvector_search_agent
-test_rag_agent
+test_rag_agent
+hybrid_search_agent

use-cases/agent-factory-with-subagents/CLAUDE.md

@@ -551,7 +551,7 @@ Build a library of common patterns:
 - ✅ Document everything
 - ✅ Validate against requirements
 
-### NEVER:
+### Anti-patterns to ALWAYS avoid:
 - ❌ Hardcode API keys or secrets
 - ❌ Skip testing phase
 - ❌ Ignore error handling

use-cases/agent-factory-with-subagents/SAMPLE_PROMPT.md

@@ -0,0 +1 @@
+I want to build a Pydantic AI agent that has the ability to perform semantic similarity search with PGVector and hybrid search. You'll need to copy over (with the cp command, you have to actually copy over everything and keep the same folder structure) everything in examples/rag_pipeline because the RAG pipeline is already set up. Look at sql/schema.sql to understand the database structure and the functions for regular and hybrid search. You'll want to use asyncpg to create the database connection and have an environment variable in .env.example for the DATABASE_URL. Build a nice CLI for the agent as well like we have in main_agent_reference. Use Archon for the Pydantic AI documentation + main_agent_reference to guide your implementation. Keep the agent simple - minimal tools, a single LLM determined by an environment variable you put in .env.example, include just the functionality that is crucial for the agent to be powerful but concise.