Add compare_pipelines skill

openclaw/skills/compare_pipelines/SKILL.md

@@ -0,0 +1,20 @@
+# compare_pipelines
+
+Run a query through both Baseline RAG and GraphRAG pipelines simultaneously, then compare their answers side-by-side with metrics including tokens, latency, cost, and answer quality.
+
+## Parameters
+- `query` (string, required): Question to compare across pipelines
+- `provider` (string, optional, default="anthropic"): LLM provider to use
+- `model` (string, optional): Specific model ID (defaults to provider's default)
+
+## Returns
+JSON with:
+- `baseline`: Answer, tokens, latency, cost from Pipeline A (Baseline RAG)
+- `graphrag`: Answer, tokens, latency, cost, entities, relations from Pipeline B (GraphRAG)
+- `complexity`: Query complexity score (0.0-1.0)
+- `recommended`: Which pipeline the adaptive router recommends
+
+## Example
+```
+compare_pipelines "Which magazine was started first?" --provider ollama --model llama3.2
+```