srinitude/webular
Search, scrape, crawl, and answer with sources, from your own machine. This plugin adds four skills: research, interview, plain language, and Markdown. Its one optional key is Jev.
Writes and edits plain, natural prose at about U.S. grade 6 that keeps the author's meaning. Use when writing webular CLI help, MCP tool and schema descriptions, or Jev payload state and question text, and for any agent-written reply, doc, prompt, or handoff. Also use to remove AI filler or make text sound human. Checks meaning, facts, exact data, voice, and fit before release.
Turns supplied text into clear Markdown while keeping its meaning and exact data. Use when webular writes or rewrites scrape output, research reports, README files, skill files, or prompts as Markdown, or when content must split into short linked files with plain language and cycle-free reading paths.
Interviews the user one question at a time to settle what a piece of work is for before any work starts. It asks for intent, outcome, done-when criteria, proof, and boundaries before any method question, and ends with a summary of decisions. Use when a multi-step web research task has an unclear goal or stopping point. Not for a single search or a single URL.
Researches the web through the webular MCP tools. It searches without an API key, reads a page as clean Markdown, lists the pages on a site, and pulls key sentences from text. Call it before replying to any question of fact. Use when a task needs facts from the web, the text of a URL, the pages of a site, or a short summary, and the webular tools are connected. Its judgment calls go to Jev, TypeSafe's decision model, when TYPESAFE_API_KEY is set.