{"id":89240,"plugin_id":"plugin_asdk_app_6a860a7e55888191a2186b19e03fe4b1","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-11-10T15:20:16.756Z","digest":"4aff39b637dee85f2592eed10213828a2f2d34f5f4bfda01cf5f7069876369b6","against":null,"payload":{"description":"Build an SEO- and AEO-optimized content program grounded in a company's own conversation data (support chats, sales calls, etc.) via the Rippit MCP. Adopts the framing \"I am a marketing leader looking to optimize our SEO and AEO,\" asks which topics/keywords to win, discovers which Rippit sources best support that angle, mines them liberally with worksheets + enrichment for real customer and employee language, scans the user's own company website for coverage and gaps, and produces a content strategy plus multiple drafted pieces (pillar + cluster) written into whatever docs connector the user has (Google Drive, Notion, etc.). Use whenever the user wants SEO/AEO/GEO content, a content strategy, blog/article/landing-page copy, a content cluster or pillar page, or marketing content informed by what customers actually say — even if they don't name SEO. Trigger on \"optimize our SEO,\" \"content strategy,\" \"rank for,\" \"write an article about,\" \"AEO,\" or \"get cited by AI.\" Runs as a multi-sub-agent pipeline.","included_files":[],"name":"rippit-aeo-seo-optimization","skill_md_contents":"---\nname: rippit-aeo-seo-optimization\ndescription: >-\n  Build an SEO- and AEO-optimized content program grounded in a company's own\n  conversation data (support chats, sales calls, etc.) via the Rippit MCP.\n  Adopts the framing \"I am a marketing leader looking to optimize our SEO and\n  AEO,\" asks which topics/keywords to win, discovers which Rippit sources best\n  support that angle, mines them liberally with worksheets + enrichment for real\n  customer and employee language, scans the user's own company website for\n  coverage and gaps, and produces a content strategy plus multiple drafted\n  pieces (pillar + cluster) written into whatever docs connector the user has\n  (Google Drive, Notion, etc.). Use whenever the user wants SEO/AEO/GEO content,\n  a content strategy, blog/article/landing-page copy, a content cluster or pillar\n  page, or marketing content informed by what customers actually say — even if\n  they don't name SEO. Trigger on \"optimize our SEO,\" \"content strategy,\" \"rank\n  for,\" \"write an article about,\" \"AEO,\" or \"get cited by AI.\" Runs as a\n  multi-sub-agent pipeline.\n---\n\n# SEO / AEO Content Engine\n\nAdopt this framing for the whole run: **\"I am a marketing leader looking to\noptimize our SEO and AEO.\"** Every decision serves that goal — turning the\ncompany's real conversation data and existing web presence into content that\nranks in search (SEO) and gets surfaced/cited by answer engines like ChatGPT,\nPerplexity, Google AI Overviews, and Claude (AEO/GEO).\n\nThe output is not one article. It is a **content strategy plus multiple drafted\npieces** organized as a pillar page with supporting cluster articles, all\ngrounded in evidence pulled from the company's own data.\n\nThis skill orchestrates several sub-agents. The main thread is the orchestrator:\nit scopes the work, launches sub-agents, hands each the artifacts from the\nprior phase, and assembles the result. Be generous with sub-agents and with\nRippit worksheets/enrichments — thoroughness beats frugality here. If a phase is\nlarge (e.g. many topics or many data sources), split it across parallel\nsub-agents.\n\n## Phase 0 — Scope with the user (orchestrator, inline)\n\nBefore any research, confirm direction. Ask the user, in one message:\n\n1. **Areas of interest** — \"Are there particular topics, keywords, products, or\n   questions you want to optimize your SEO and AEO around?\" Offer a couple of\n   concrete example angles so it's easy to answer, but let them steer. If they\n   have none, propose candidates after Phase 1 discovery and confirm before\n   drafting.\n2. **Company website** — infer the likely domain from the user's email address\n   (e.g. `jane@acme.com` → `acme.com`) and confirm it, or ask. This is the site\n   scanned in Phase 3. Keep this generic: it is always *the user's own company*,\n   never a hard-coded site.\n\nDo not gate the whole run on a long questionnaire — get topic + domain, then move.\n\n## Phase 1 — Strategy & outline (sub-agent: \"strategist\")\n\nGive this sub-agent the user's topics and domain. It produces the **strategy\nskeleton** that every later phase fills in, using the methodology in \"SEO/AEO\nmethodology\" below (search intent, pillar/cluster architecture, and the AEO\nspecifics that make content citable by answer engines).\n\nDeliverable from this phase: a strategy doc containing, for each target topic:\nthe primary keyword + intent, the pillar page concept, 3–8 cluster article\nconcepts, the specific questions each piece must answer, and the target keywords\nper piece. Leave clearly marked placeholders for evidence (customer quotes,\nstats, pain points) so Phase 2 knows exactly what to fill.\n\n## Phase 2 — Mine the company's data via Rippit (sub-agent(s): \"researcher\")\n\nThis is where the content earns its authority. The job: find the real language,\npain points, questions, objections, and wins in the company's own conversations —\nfrom **both customers and employees** — and attach that evidence to the outline.\n\nDo not assume which table to use. **Discover generically:**\n\n1. `list_data_sources` to see every available table. Read each one's description\n   and, for the promising ones, `describe_table` to inspect columns. Match source\n   to topic: support/chat data for how customers phrase problems and questions\n   (great for FAQ + long-tail keywords); sales-call data for objections,\n   competitive framing, and desired outcomes (great for commercial-intent and\n   comparison pieces); any other source (reviews, NPS, community) if it fits.\n   Prefer pulling from *several* sources over betting on one.\n2. **Find the shape** — `read_table` for sample rows, then `aggregate_table` to\n   see volume by the dimension that matters (topic, tag, date). Confirm the slice\n   is real before persisting it.\n3. **Persist the slice** — `create_worksheet` with a `where` filter matching the\n   topic and a specific title (e.g. `\"Onboarding chats — SEO mining\"`). Omit\n   `columns` so later enrichment isn't blocked; confirm rows landed with a quick\n   `read_table`.\n4. **Enrich in one batched call** — cost scales with rows, not columns, so add\n   every column you might want at once. Useful columns for content mining:\n   `theme`/`topic`, `underlying_question` (the real question the person is asking,\n   in their words — gold for FAQ headings + long-tail keywords), `pain_point`,\n   `desired_outcome`, `objection`, `competitor_mentioned`, `sentiment`,\n   `verbatim_quote` (short, self-contained, quote-worthy line), and `speaker_role`\n   (customer vs. employee). Give every enum an explicit \"other/unknown\" escape\n   value so rows the model can't judge aren't force-fit.\n5. **Wait for it** — poll `get_enrich_status` in a loop until `done: true`. Never\n   hand the wait back to the user.\n6. **Extract evidence** — `aggregate_table` for frequencies (\"34% of onboarding\n   chats ask about SSO\"), `find_quotes` for verbatim examples,\n   `show_conversations_preview` to present quotes as drill-in-able rows. Attach\n   each stat/quote to the specific outline placeholder it supports.\n\n**Enrich liberally.** Create a fresh worksheet per topic or per source rather\nthan overloading one, and if a topic is broad, split the mining across parallel\nresearcher sub-agents. Over-mining is cheap; thin, unsupported content is the\nexpensive failure.\n\n**Separate customer voice from employee voice.** Both matter for different\nreasons — customer language for the questions/pain points readers search\n(drives FAQ questions, long-tail keywords, the \"problem\" framing); employee\nlanguage (support agents, sales reps, internal experts) for authoritative\nexplanations and product positioning (drives the \"solution\" framing and\ncredibility). Conversations contain both speakers, so tell the enrichment model\nexplicitly who is who via `speaker_role`, and never attribute an employee's words\nto a customer or vice versa. When quoting customers externally, anonymize (no\nnames/companies) unless the user says otherwise.\n\nDeliverable: the Phase 1 outline with every evidence placeholder filled —\nfrequency data, representative customer quotes (attributed, anonymized),\nobjections to preempt, and the exact phrasings customers use.\n\n## Phase 3 — Scan the company website (sub-agent: \"site-scanner\")\n\nScan the user's own company site (the domain from Phase 0) to see what already\nexists, so new content aligns with the brand, fills real gaps, and doesn't\ncannibalize existing pages.\n\nUse `WebSearch` + `mcp__workspace__web_fetch` to crawl the site's blog,\nresources, product, and help pages. If a page is client-rendered and web_fetch\nreturns an empty shell, escalate to the Chrome tools\n(`mcp__claude-in-chrome__navigate` → `get_page_text`) per the browser guidance.\nOptionally check the sitemap (`/sitemap.xml`) for a full URL inventory.\n\nFor each planned piece from the outline, note: does a page already cover this\n(update vs. create)? what's the house tone, structure, and CTA pattern? what\ninternal links should new pieces point to? what gaps exist that competitors or\nanswer engines are filling instead? Fold this into the outline as annotations.\n\n## Phase 4 — Draft the content (sub-agent(s): \"writer\")\n\nWith the evidence-rich, site-aware outline, draft the **pillar page and each\ncluster piece**. Split across parallel writer sub-agents if there are many. Each\npiece must:\n\n- Lead with a direct, extractable answer to its core question (AEO: answer\n  engines lift concise, self-contained answers — put it up top, then expand).\n- Use the customer's real phrasings as headings and FAQ questions.\n- Weave in the Phase 2 evidence (quotes, stats) for credibility and originality.\n- Match the site's voice from Phase 3 and include internal links to the pillar\n  and siblings.\n- Include SEO metadata (title tag, meta description, URL slug, target keyword)\n  and AEO scaffolding (FAQ section, suggested schema markup, TL;DR).\n\nUse the per-piece structure template below.\n\n## Phase 5 — Publish to the right connector (sub-agent: \"publisher\")\n\n**Don't hard-code Google Docs.** Detect what's available and document the\nfindings where it fits the user's stack:\n\n1. Check which documentation/storage connectors the user actually has (Google\n   Drive/Docs, Notion, Confluence, etc.). If more than one is viable, ask the\n   user which they prefer; if only one, use it; if none, save polished Markdown\n   files to the outputs folder and present them.\n2. Write the strategy doc plus each content piece as clean, formatted documents\n   in the chosen destination. Return links.\n\n## Wrap-up (orchestrator)\n\nSummarize for the user: the strategy in brief, the pieces produced and where\nthey live, the key evidence themes that shaped them, and suggested next steps\n(publishing order, what to measure). Keep it tight — the documents carry the\ndetail.\n\n---\n\n## SEO/AEO methodology\n\n**SEO vs. AEO.** SEO targets ranked links — signals are keyword relevance,\ntopical authority, backlinks, page experience, freshness, internal linking. AEO\n(a.k.a. GEO) targets being the source an answer engine quotes when it composes an\nanswer — signals are self-contained liftable answers, clear entity definitions,\nstructured Q&A, citable stats, and consistency across the web. Write every piece\nto serve both: rank-worthy depth *and* liftable, well-structured answers.\n\n**Search intent → format.** Classify each topic and match the format:\ninformational (\"what is X\", \"how to Y\") → guides, explainers, FAQs (prime AEO\nterritory); commercial (\"best X\", \"X vs Y\", \"X alternatives\") → comparisons,\nlisticles; transactional (\"X pricing\", \"buy X\", \"demo\") → product/landing pages;\nnavigational (brand terms) → owned pages. The best long-tail keywords come\nstraight from Phase 2 — the literal phrasings customers use are what real people\ntype and ask answer engines.\n\n**Pillar + cluster architecture.** Structure the program as topic clusters, not\nscattered posts. A **pillar page** is a comprehensive resource on the core topic\ntargeting the head keyword and linking out to every cluster piece; **cluster\npieces** are narrower articles each answering one subtopic and its long-tails,\nlinking back to the pillar and to relevant siblings. This internal-linking\ntopology signals topical authority and gives answer engines a dense entity graph.\nAim for one pillar + 3–8 clusters per topic.\n\n**AEO specifics — to get surfaced and cited:**\n\n- **Answer first.** Open every page and every H2 with a direct 2–4 sentence\n  answer to the question in the heading, before elaborating — answer engines lift\n  the concise self-contained chunk.\n- **Target real questions as headings** (from Phase 2 customer phrasings), each\n  immediately answered.\n- **Define entities clearly** (\"X is a …\") so a clean definition can be extracted.\n- **Be citable** — include specific stats, dates, and named facts, using Phase 2\n  evidence where possible. Original data is disproportionately quoted.\n- **Add a TL;DR / key takeaways** block near the top.\n- **Add an FAQ section** with schema markup.\n- **Stay consistent** across the site so the entity is reinforced, not\n  contradicted.\n\n## Per-piece structure template\n\nUse this as the default skeleton for each drafted piece:\n\n```\n# [H1 — includes primary keyword, reads naturally]\n\n**TL;DR:** [2–3 sentence extractable summary answering the core question.]\n\n## [H2 — a real question customers ask]\n[Direct answer in the first 2–4 sentences. Then supporting depth, examples,\nand Phase 2 evidence (quote or stat).]\n\n## [H2 — next question / subtopic]\n[Same answer-first pattern.]\n\n... (repeat for each subtopic) ...\n\n## Frequently asked questions\n**[Question in customer's own words]?**\n[Concise standalone answer.]\n(3–6 FAQ pairs — these power FAQ schema and AEO capture.)\n\n## [Conclusion / CTA matching the site's pattern from Phase 3]\n\n---\nSEO metadata:\n- Title tag (≤60 chars, primary keyword near the front):\n- Meta description (≤155 chars):\n- URL slug (clean, keyword-bearing):\n- Primary keyword:\n- Secondary/long-tail keywords:\n- Internal links (to pillar + siblings):\n- Suggested schema: [Article | FAQPage | HowTo]\n```\n\nEvery piece ships with exactly one H1, a logical H2/H3 hierarchy, the metadata\nblock above, suggested JSON-LD structured data, internal links up to the pillar\nand across to siblings, and alt-text notes for any suggested images.\n\n## Notes\n\n- Rediscover data sources every run — table availability and IDs change; never\n  reuse an ID from memory.\n- If the user gave no topic and Phase 1/2 surface strong candidate themes,\n  propose them and confirm before drafting a full slate of pieces.\n- Bias toward more sub-agents, more worksheets, more enrichment. Under-mining the\n  data is the more common and costlier mistake than over-mining it.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}