Menu
Web search using DuckDuckGo (primary) and Brave Search (fallback) via the web-browser skill. Use for searching documentation, facts, or any web content when API-based search is not available.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Allows to interact with web pages by performing actions such as clicking buttons, filling out forms, and navigating links. It works by remote controlling Google Chrome or Chromium browsers using the Chrome DevTools Protocol (CDP). When Claude needs to browse the web, it can use this skill to do so.
Fetch up-to-date, version-specific documentation for any library, framework, or tool from Context7 (context7.com) into context. Use whenever accurate API references, installation steps, or usage examples are needed instead of relying on training-data knowledge that may be outdated — e.g. "how do I bootstrap Laravel 13", "current options for X", "the correct signature for Y in version Z".
Converts a Markdown file to a professionally styled PDF with cover page, hierarchical headings, tables, links, blockquote callouts, and proper page breaks. Use when the user wants a polished PDF version of a markdown document (research reports, documentation, READMEs, briefs). Output is rendered via headless Chrome from a styled HTML intermediate.
Generates per-job-tailored ATS-friendly resumes and cover letters. Produces single-column HTML, PDF (via Chrome CDP), and DOCX (via python-docx). Reads a user-maintained content database for experiences/skills; extracts keywords from a target job description; outputs are named by {date}-{company}-{role}. Use when the user pastes a job description (or URL) and wants a tailored application package. Do NOT use for the dark-sidebar visual resumes (those are the existing chrisli-resume*.html files).
Renders one or more HTML files to PDF using Chrome's headless --print-to-pdf mode. Reliable and standalone — no pre-running Chrome or CDP connection required, and a throwaway user-data-dir means it never collides with the user's running browser. Use for ATS resumes, cover letters, or any local HTML→PDF conversion. By default it forces A4 with 10mm margins on every page; headers/footers are omitted.
Extract and analyze content from PDF files using vision AI. Supports single or multiple PDFs (batch), full text extraction, structured data extraction with Pydantic schemas, Markdown/HTML conversion, worksheet-clean modes, user-content extraction, and custom prompt extraction. Use when you need to extract, summarize, or analyze content from one or more PDF documents.
| name | web-search |
| description | Web search using DuckDuckGo (primary) and Brave Search (fallback) via the web-browser skill. Use for searching documentation, facts, or any web content when API-based search is not available. |
Interactive web search using the web-browser skill. Searches DuckDuckGo by default, falls back to Brave Search if needed.
Note: This skill does NOT use Google for any searches.
This skill drives a single shared Chrome tab through the web-browser skill. Running multiple search.js invocations concurrently (e.g. as parallel bash tool calls) will cause them to race for the same tab and return cross-contaminated results — each invocation may extract whatever the previous one left on the page.
search.js enforces this with a file lock: a concurrent invocation will exit immediately with an error rather than produce garbage output.
Run them sequentially in separate tool calls and wait for each result before issuing the next. The recommended pattern is:
node .../search.js "first query" -n 5
# wait for output
node .../search.js "second query" -n 5
# wait for output
node .../search.js "third query" -n 5
If you genuinely need to script a batch sequentially, use the helper:
node .../scripts/batch-search.js "query one" "query two" "query three" -n 5
This runs the queries one at a time against the same shared Chrome and emits a clearly delimited result block per query. Never wrap the per-query search.js in a shell parallelizer like xargs -P or parallel.
Requires the web-browser skill. Chrome is auto-started if not already running — no manual step needed.
If auto-start fails, you can start Chrome manually:
{baseDir}/../web-browser/scripts/start.js # Fresh profile
{baseDir}/../web-browser/scripts/start.js --profile # Copy your profile (cookies, logins)
{baseDir}/../web-browser/scripts/start.js --headless # No GUI needed
{baseDir}/scripts/search.js "query" # Basic search (5 results)
{baseDir}/scripts/search.js "query" -n 10 # More results
{baseDir}/scripts/search.js "query" --brave # Use Brave Search directly
Reminder: one query at a time, per the warning above.
-n <num> - Number of results (default: 5)--brave - Use Brave Search instead of DuckDuckGo--- Result 1 ---
Title: Page Title
URL: https://example.com/page
Snippet: Description from search results
--- Result 2 ---
...
--brave flag used)