Ranksmile

One action of this workspace with what the user sees: overview, why (with its numbers), steps and which are done, the evidence read on the pages, the pages and prompts involved, the competitor pages behind it, and the content brief for owned actions.

Recommended actions for this workspace, highest impact first: goal site (fix the site), owned (write or improve a page), earned (get onto an outside page: surface editorial, ugc or reference) or claims (correct what answers say). Each with impact (1-5), status (new, in_progress, done, declined), topic and steps done. actions__get has the details.

Get one article: metadata, stored scores, and (optionally) the HTML body. Read the body when judging writing quality; skip it when you only need numbers.

The most recent analysis / generation jobs for an article, with status, payload and result. Use this when generated content looks wrong: it shows what the generator was told and what it returned. Payloads and results over 4000 characters are cut to a prefix.

List articles with their stored scores and word counts. Filter by workspace_id or status. Use this first to find the article ids the other tools take.

The Auto-Optimize / publish history for an article: before and after score, before and after length, and the run metadata (mode, phase, rejection reason). Use this when Auto-Optimize appears to change nothing.

Explain an article's content score: the per-slot breakdown, word/heading/paragraph counts against their competitor-derived targets, and every NLP term with its current vs. required usage. This is the tool that says WHY a generated article scores badly.

The last site crawl for this workspace: site health and AI-search health (0-100), pages crawled, error and warning counts, and each issue with how many pages it affects, most severe first. Reads the stored crawl; starts nothing.

The last 30 days of Google Search Console for this workspace's site: clicks, impressions, CTR (%) and average position per query or per page, most clicks first. Pages or queries with many impressions at position 8-20 are the quickest wins.

Google positions of the keywords this workspace tracks (checked daily): current position (null = not in the checked results), best position ever found, the ranking URL and search volume.

How often AI answers (ChatGPT, Perplexity, Gemini, Google AI) name this workspace's brand in the latest scan: the brand score, the brands named instead, the score per topic, the trend over the last scans, and every tracked prompt with its intent, tags, per-engine score and mention rate, least-mentioned first. The prompts at 0 are the gaps content can close; visibility__prompt shows why one is lost. The site-wide parts (overview, competitors, topics, history, scans_total, unanswered_prompts) come on the first page only; later pages carry the next prompts.

One tracked prompt in the latest scan: each engine's answer, the brands it names (with sentiment and quotes), the pages it cites, the sub-queries the engine searched (fan-out), and the stored "how to get cited" advice. Use it to see why a prompt is lost, which pages to get onto, and how buyers phrase the question.

The pages (by: "url") or sites (by: "domain") the latest scan's AI answers cite, most cited first, each with its type (UGC, Editorial, Reference, Institutional, Corporate, Competitor, You, Other). Per page: whether the answers citing it name the brand, whether the page itself does, and the brands named instead. gap: true marks pages that help competitors and not the brand: Editorial ones are outreach targets, UGC ones community threads, Competitor ones show what to out-write.

What Ranksmile knows about a workspace: the brand name and notes the editor writes from, the voices, the competitors found in search, and the research log (the newest 20 entries of the last 90 days, with age_days). Call it before SEO work. Before paying for research, reuse a log entry under 30 days old. missing lists what is still empty.

List the Ranksmile workspaces the connected account can access, with their primary domain.

Save what you learned so the editor and other agents use it. Ops apply in order; if one is invalid, none is applied. To ADD a fact to the brand notes use {"append_brand_knowledge": "..."}: it keeps everything already there. {"brand_knowledge": "..."} REPLACES the whole notes ("" clears them): use it only to rewrite or shorten, after reading them with workspace__get_context. The notes hold at most 4000 characters. {"append_research_log": ": . Verdict: "} records research you did (max 500 chars). Confirm facts with the user before storing them. Returns the updated context.