Add CITATION.cff

CITATION.cff

@@ -0,0 +1,20 @@
+cff-version: 1.2.0
+message: "If you use benchflow in your research, please cite it as below."
+title: "BenchFlow: framework for RL environments for LLM agents"
+abstract: "BenchFlow is a framework for building RL environments to evaluate and train LLM agents. Built on the Agent Client Protocol (ACP), it provides Scene-based multi-turn, multi-agent, and multi-model evaluation in shared sandboxes — without Docker Compose or sidecar containers. Supported use cases include interactive user simulation, code-review loops, bring-your-own-skill (BYOS) skill generation, multi-turn iterative refinement, cross-model review (cheap coder + strong reviewer), and stateful service tasks against live mock APIs (Gmail, Calendar, Docs, Drive, Slack). See docs/use-cases.md."
+type: software
+authors:
+  - name: "BenchFlow team"
+    website: "https://github.com/benchflow-ai/benchflow"
+repository-code: "https://github.com/benchflow-ai/benchflow"
+url: "https://github.com/benchflow-ai/benchflow"
+license: Apache-2.0
+version: 0.3.2
+keywords:
+  - benchmark
+  - llm-agents
+  - acp
+  - agent-evaluation
+  - multi-turn
+  - terminal-bench
+  - skillsbench