feat: add llm.txt — agent entry point for core/agent

Standard llm.txt with package layout, test coverage stats, conventions.
Points to CLAUDE.md and RFC-025 for full specs.

Co-Authored-By: Virgil <virgil@lethean.io>

llm.txt

@@ -0,0 +1,32 @@
+# core/agent — AI Agent Orchestration
+
+> dappco.re/go/agent — Dispatch, monitor, and verify AI agents (Codex, Claude,
+> Gemini) working on sandboxed repositories. MCP server for agent coordination.
+
+## Entry Points
+
+- CLAUDE.md — Agent instructions, build commands, architecture overview
+- pkg/agentic/ — Core orchestration (dispatch, prep, verify, scan, review)
+
+## Package Layout
+
+- cmd/core-agent/ — CLI entry point (97 lines — just core.New + services)
+- pkg/agentic/ — Agent orchestration: dispatch, workspace prep, PR creation, verification
+- pkg/brain/ — OpenBrain memory integration
+- pkg/lib/ — Embedded templates, personas, flows, workspace scaffolds
+- pkg/messages/ — Typed IPC message definitions (12 message types)
+- pkg/monitor/ — Agent monitoring, notifications, completion tracking
+- pkg/setup/ — Workspace detection and scaffolding
+
+## Test Coverage
+
+- 840 tests, 79.9% statement coverage
+- 92% AX-7 (Good/Bad/Ugly) category coverage
+- 14 functions awaiting go-process v0.7.0 for full testability
+- Zero os/exec imports in source — all commands via go-process
+
+## Conventions
+
+Follows RFC-025 Agent Experience (AX) principles.
+Test naming: TestFile_Function_{Good,Bad,Ugly}
+See: https://core.help/specs/RFC-025-AGENT-EXPERIENCE/