diff --git a/skills/data-workflow-design/SKILL.md b/skills/data-workflow-design/SKILL.md
new file mode 100644
index 0000000..eef3039
--- /dev/null
+++ b/skills/data-workflow-design/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: data-workflow-design
+description: "Design reusable extraction and transformation workflows for CSV, Excel, JSON, JSON-LD, and other structured data using a target JSON Schema."
+---
+# data-workflow-design
+
+# Data Workflow Design
+
+When creating an extraction and transformation workflow:
+
+1. Identify the source format and parse strategy (csv, excel, json, jsonld, or text).
+2. Inspect a bounded sample only; never assume unbounded rows are available.
+3. Treat the supplied JSON Schema as the target contract. Required fields must be mapped or listed as gaps.
+4. Prefer exact source field matches first, then case/space/punctuation-normalized matches.
+5. Add transformation steps for type coercion, enum checks, date normalization, number parsing, trimming, and default handling.
+6. Include validation gates before output: required fields, type checks, additionalProperties behavior, and per-field constraints.
+7. Save a reusable JSON workflow spec with version, source profile, target schema, mapping rules, validation rules, and an execution receipt reference.
+8. Never include credentials, platform tokens, database URLs, or raw oversized data in saved workflows.