novada-agent

Novada Agent Skill

You have access to Novada's 37 curated MCP tools (the set novada_discover lists; the hosted endpoint exposes a core-derived 29-tool subset). This skill tells you exactly which tool to use when and how to use it effectively. Older names like novada_unblock, novada_verify, and novada_health still work as back-compat aliases — the current tools are shown below.

Deep-dive companion skills: novada-scrape (platform scrapers + price fields), novada-extract (single/batch extraction), novada-browser (CDP automation), novada-proxy (proxy type selection + escalation), novada-site-copy (whole-site copying / RAG ingestion).

Tool Selection — Decision Tree

Need web data?
├── Have specific URLs already? → novada_extract
├── Need to find URLs first?
│   ├── Question needs facts/prices/current events (accuracy matters)?
│   │   ├── 1 source → novada_search THEN novada_extract top result (read it)
│   │   └── Multiple sources → novada_research (reads full content for you — cheaper than N search+extract)
│   ├── Just locating pages (no answer needed yet) → novada_search
│   ├── Explore an entire site → novada_map or novada_crawl
│   └── Multi-faceted question → novada_research
└── Want ranked, cited source material to reason over? → novada_research

Need structured platform data (Amazon, TikTok, LinkedIn…)? → novada_scrape_<platform> if one of the 15 dedicated platform tools exists for it, else generic novada_scrape
Need a proxy config template (endpoint + masked credentials) for your own HTTP requests? → novada_proxy
Page blocked or JS-heavy, need raw HTML? → novada_extract (format:"html", render:"render")
Need to click/fill/screenshot a page? → novada_browser
Check account balance / plans / entitlements? → novada_account

Escalation ladder (pick the lowest rung that answers the question): novada_search (locate) → novada_extract top result (read one page) → novada_research (read many sources)

Also read novada://guide — it contains the full decision tree and workflow patterns.