# Content Desk > Paste a keyword export (Ahrefs, Semrush, Search Console, Keyword Planner or a plain list), what customers actually ask (sales-call questions, support tickets, objections, survey answers) and what your company does. The browser reads every keyword's buyer stage and search intent, groups topic clusters, finds quick wins and content gaps, matches the research to the clusters and splits the calendar 60/30/10. A paid plan lane picks 3-5 content pillars and scores 5-10 priority topics on the skill's four weighted factors; a paid piece lane plans one flagship piece with ten titles, an outline and its distribution. URL: https://content-desk.skillsafe.ai/ API tutorial: https://content-desk.skillsafe.ai/api.html ## Who it is for Founders and marketers at software and service companies who have a keyword tool export and a pile of customer questions and need to decide what content to make next: which 3 to 5 topics to own, which pieces to make first, and how to plan the first one so it is found, shared and leads to the product. ## The free scan (in the browser, no model) - Keyword exports with a header row (Keyword, Volume or Avg. monthly searches or Impressions, KD or Difficulty or Competition, Position, URL), CSV, TSV or semicolon-separated, or one keyword per line ("invoice template, 2400, 18"). Search Console impressions stand in for volume and are labelled; Keyword Planner ranges read as their lower bound; duplicates and rows without a keyword are skipped and counted. - Buyer stage from the skill's modifiers: awareness (what is, how to, guide to, introduction to), consideration (best, top, vs, alternatives, comparison), decision (pricing, reviews, demo, trial, buy), implementation (templates, examples, tutorial, how to use, setup); intent informational, commercial or transactional; a content-type hint. - Topic clusters by shared term (a word in over 60% of the keywords is the category and makes no cluster), volume per stage and per cluster. - Quick-win candidates: difficulty 30 or under (or low competition) at or above the export's median volume. Content gaps: keywords with no position or below 20, when the export has positions. - Customer research split into items (question, objection, pain, note, with the source when labelled) and matched to clusters; 30% or more of items naming a cluster's term marks a theme. - Publishing capacity ("4 posts a month") and the 60/30/10 split into searchable, shareable and experimental pieces. - Flags CS-01..CS-12: no keywords, no research, no volume, no difficulty, one keyword holding most of the volume, no consideration or decision keywords, no goal, no capacity, no competitors, rows skipped, most keywords unclassified, no positions. - Exports: the scan as Markdown and the full keyword map as CSV. ## The paid lanes (one system prompt, routed on `task`) - `plan`: the goal; 3-5 pillars, each product-led, audience-led, search-led or competitor-led, quoted from the input, with the page's cluster ids; 5-10 priority topics (searchable, shareable or both; content type; the export's keyword id and exact keyword; buyer stage; the research ids behind it) scored 1-10 on customer impact, content-market fit, search potential and resources with the total weighted 40/30/20/10, highest first; up to 5 quick wins from the page's candidates; the calendar split; a hub-and-spoke cluster map. - `piece`: ten title options and the pick; kind, content type, keyword, buyer stage, intent; a slug under /blog/; the angle and the first line (the keyword in title, first line and slug for a searchable piece); a 4-8 section outline; the evidence it needs as [placeholders]; whether the format earns links, from the skill's link-earning table; the five editing passes; owned, rented and borrowed distribution; lines that stand alone as posts. A plan topic hands over to this lane with one button. Every reply is checked against the input: quotes word for word, every keyword, cluster and research id the page's own, each keyword's exact text and stage, the weighted totals and their order, the score caps, the quick-win source, the calendar sum, the title count and pick, the editing passes, an owned channel, and no invented volume, count, percentage, date or duration. ## Source Derived from the agent skill @coreyhaines31/content-strategy (https://skillsafe.ai/skill/@coreyhaines31/content-strategy), by Corey Haines, in coreyhaines31/marketingskills (MIT): https://github.com/coreyhaines31/marketingskills/tree/main/skills/content-strategy - searchable versus shareable content, the content types, link-earning formats, content pillars and topic clusters, keyword research by buyer stage, the content ideation sources, the four-factor prioritization score, the 60/30/10 calendar split, per-format execution discipline and Create Once, Distribute Twice; and its reference references/content-distribution.md (the ORB framework: owned, rented, borrowed). Runs on gpt-terra. Part of SkillSafe (https://skillsafe.ai/).