diff --git a/ddgs-search/SKILL.md b/ddgs-search/SKILL.md
new file mode 100644
index 0000000..88d449b
--- /dev/null
+++ b/ddgs-search/SKILL.md
@@ -0,0 +1,82 @@
+---
+name: ddgs-search
+description: >
+  Web metasearch via DDGS (Dux Distributed Global Search). Aggregates results from
+  multiple search engines (Bing, Brave, DuckDuckGo, Google, Yandex, Yahoo, and more).
+  Use for any web search need during investigation.
+---
+
+# Identity
+
+You have access to a metasearch daemon reachable via `web_fetch`. Use it freely
+whenever you need to find information on the web.
+
+---
+
+# How to Search
+
+## Text search
+
+```
+web_fetch(url="http://192.168.100.203:8000/search/text?query=QUERY")
+```
+
+## News search
+
+```
+web_fetch(url="http://192.168.100.203:8000/search/news?query=QUERY")
+```
+
+URL-encode the query — use `+` for spaces (e.g. `query=Bitcoin+whitepaper`).
+
+---
+
+# Parameters
+
+All parameters are optional except `query`.
+
+| Parameter    | Default    | Description                                      |
+|--------------|------------|--------------------------------------------------|
+| `query`      | —          | Your search query (required)                     |
+| `max_results`| 10         | Number of results to return                      |
+| `timelimit`  | none       | `d` (day), `w` (week), `m` (month), `y` (year)  |
+| `backend`    | auto       | Round-robins across all available engines        |
+| `region`     | us-en      | Region/language code                             |
+| `safesearch` | moderate   | `on`, `moderate`, `off`                          |
+| `page`       | 1          | Page of results                                  |
+
+---
+
+# Response
+
+```json
+{
+  "results": [
+    {
+      "title": "...",
+      "href": "...",
+      "body": "..."
+    }
+  ]
+}
+```
+
+---
+
+# Usage Pattern
+
+Search first, then fetch full content for the most relevant results:
+
+1. **Search** — call the endpoint to get titles, URLs, and snippets
+2. **Pick** — identify the most relevant URLs
+3. **Fetch** — use `web_fetch` on those URLs to read full page content
+
+---
+
+# Tips
+
+- Whitepapers: `query=<project>+whitepaper+filetype:pdf`
+- Audits: `query=<project>+security+audit`
+- Recent news: use `/search/news` with `timelimit=m`
+- Team info: `query=<project>+team+founders`
+- Community: `query=<project>+reddit` or `query=<project>+bitcointalk`
\ No newline at end of file