expandai 0.1.6 → 0.1.7

package/build/skills/expandai/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: expandai
-description: Use when any public http(s) URL needs to be read, including documentation, articles, GitHub pages, blog posts, release notes, or a link the user pasted, and before calling WebFetch or curl on it. WebFetch does not execute JavaScript, so also use this whenever a fetch came back empty, truncated, 403, paywalled, or asking you to enable JavaScript. Cite every claim from captured pages using evidence IDs and capture URLs; read the Citations section before answering. Markdown may contain `[appendix omitted: … ] {N}` markers where navigation, footers and related links were moved to the appendix; resolve them with `expandai appendix` rather than refetching. Not for localhost, private or internal hosts, or JSON API endpoints.
+description: Use when any public http(s) URL needs to be read, including documentation, articles, GitHub pages, blog posts, release notes, or a link the user pasted, and before calling WebFetch or curl on it. WebFetch does not execute JavaScript, so also use this whenever a fetch came back empty, truncated, 403, paywalled, or asking you to enable JavaScript. Also use it when you need to find pages on the web and have no URL yet (`expandai search`). Cite every claim from captured pages using evidence IDs and capture URLs; read the Citations section before answering. Markdown may contain `[appendix omitted: … ] {N}` markers where navigation, footers and related links were moved to the appendix; resolve them with `expandai appendix` rather than refetching. Not for localhost, private or internal hosts, or JSON API endpoints.
 metadata:
   expandai-profile-version: 0.0.4
 ---
@@ -29,6 +29,8 @@ JavaScript engine, so on a client-rendered page it returns the empty shell and n
 Skip it for localhost, private or internal hosts, and JSON API endpoints; `curl` is the right
 tool there.
 
+When you need pages on the web but have no URL yet, run `expandai search` first. See Search below.
+
 ## Usage
 
 ```bash
@@ -237,6 +239,20 @@ Then **check that the row printed next to your number is the row you mean**. The
 scan, not a parser. That check is what catches the shapes it gets wrong: tables in blockquotes, a
 fenced block containing a shorter fence, and tables without leading pipes. Count by hand for those.
 
+## Search
+
+When you need pages but have no URL, search first, then fetch the results worth reading:
+
+```bash
+expandai search "how do I rotate a Stripe API key" --max-results 3
+expandai fetch https://docs.stripe.com/keys --highlights "how do I rotate a Stripe API key"
+```
+
+Each result has a title, a URL and an excerpt. The excerpt is a preview, not the page:
+fetch the page and cite what you take from that fetch, never from the excerpt.
+
+Text output exits non-zero when there are no results. `--format json` returns an empty `results` list instead.
+
 ## Common patterns
 
 ```bash

package/package.json

@@ -1,7 +1,7 @@
 {
 	"name": "expandai",
 	"type": "module",
-	"version": "0.1.6",
+	"version": "0.1.7",
 	"private": false,
 	"description": "Unified CLI and MCP server for ExpandAI",
 	"author": "ExpandAI",
@@ -69,7 +69,7 @@
 		"@effect/platform-node": "0.106.0",
 		"@effect/vitest": "0.29.0",
 		"@expandai/schema-utils": "file:./vendor/schema-utils",
-		"@expandai/sdk": "0.20.4",
+		"@expandai/sdk": "0.21.0",
 		"@modelcontextprotocol/sdk": "1.29.0",
 		"@openai/codex-sdk": "^0.133.0",
 		"@opencode-ai/sdk": "1.15.10",
@@ -86,5 +86,5 @@
 		"mime": "^4.1.0",
 		"ws": "^8.21.0"
 	},
-	"expandSourceSha": "7882e4eb2760e446c5ce20b83cb9a09f9017e639"
+	"expandSourceSha": "7b3661609d7d78b3d74d3ac67fb39b368e7c1aeb"
 }