← TopicLift Editorial InsightsCONTENT HISTORY

Update to TopicLift Editorial Insights

Snapshot Oct 25, 2026 · 12:04 UTC · version 1.0.1

Collection source: downloaded plugin package.

WHAT CHANGED

First saved snapshot

No earlier snapshot is available to establish a change.

Save TopicLift Editorial Insights and check future changes in Following.

Up to 3 plugins free · no card required.

Compare saved observations

Download comparison JSON
Full technical diff · 0 changed fields
Full snapshot data
{
  "description": "Use authenticated TopicLift data to answer questions about realtime and historical editorial traffic, Google Search and Discover performance, and what a publisher should publish next. Use only TopicLift MCP data for editorial recommendations. Public installations use TopicLift OAuth 2.1; local Codex and Claude development may use the documented personal token.",
  "included_files": [],
  "name": "topiclift-editorial-copilot",
  "skill_md_contents": "---\nname: topiclift-editorial-copilot\ndescription: Use authenticated TopicLift data to answer questions about realtime and historical editorial traffic, Google Search and Discover performance, and what a publisher should publish next. Use only TopicLift MCP data for editorial recommendations. Public installations use TopicLift OAuth 2.1; local Codex and Claude development may use the documented personal token.\n---\n\n# TopicLift Editorial Copilot\n\nUse the `topiclift` MCP server whenever the user asks about a TopicLift site's traffic, article performance, Google Search, Google Discover, aggregate search queries, an editorial idea, or publishing recommendations.\n\n## Operating rules\n\n- Detect the language of the current user message, not the conversation history, site language, tool output, or account locale. Write the complete conversational answer in that language: English input requires an English answer, Italian input requires an Italian answer, and so on. Keep tool names, schema fields, URLs, and quoted source labels unchanged when needed, but translate explanatory prose, headings, recommendations, titles, and angles. If the current message is genuinely mixed or ambiguous, follow its dominant language.\n- On every TopicLift tool call, set `response_language` to the BCP 47 tag for the current user message (`en` for English, `it` for Italian, and so on). Follow the returned `response_guidance` when composing the answer.\n- Use only data returned by TopicLift. Do not browse the web or add external trend data to fill gaps.\n- Keep every analysis scoped to one site. If the site is missing or ambiguous, call `list_sites` and ask the user to choose only when more than one plausible site remains.\n- Select the period from the question. Use realtime for current activity; otherwise use today, yesterday, 7 days, 28 days, or 90 days as the closest supported range.\n- Use comparisons when they materially help. The server is authoritative for metrics and comparison windows.\n- Never claim access to individual visitors or raw sessions. TopicLift exposes editorial and aggregate Search Console data only.\n- Do not modify content, settings, opportunities, or WordPress. This integration is read-only.\n- Never reveal or repeat the user's access token.\n\n## Traffic questions\n\nUse `get_traffic_overview` for totals, trends, sources, and realtime status. Its traffic metrics are restricted to synced, published editorial WordPress posts; homepage, pages, archives, unknown post IDs, and other non-editorial URLs are excluded. State this scope briefly when reporting the numbers. Use `get_top_articles` when the question is about individual articles, winners, declines, or unusual growth.\n\nAlways identify the site and period in the answer. Include the most relevant metrics and supporting articles. Do not overwhelm a short operational question with every available metric.\n\n## Editorial idea review\n\nUse `review_editorial_idea` when the user provides an idea, provisional title, outline, draft, or existing article URL and asks whether it is worth publishing or how to improve it. The tool is read-only and compares the input with the selected site's published editorial inventory and aggregate Google evidence.\n\nExplain the overlap risk, matching Search demand, related articles, confidence, and the suggested angles or titles. For a URL, distinguish an update opportunity from a genuinely new article. If the input is only an idea, do not imply that a URL or complete draft is required.\n\n## Google Search and Discover questions\n\nUse `get_google_performance` for Search, Discover, and Google News performance. Use `get_search_queries` only for aggregate web-search queries. All Google metrics, pages, queries, and associated articles returned by these tools are restricted to synced, published editorial WordPress posts with a non-empty canonical URL; non-editorial pages, homepage/archive URLs, drafts, and unmapped Search Console rows are excluded. Treat missing Discover query data as expected because Search Console does not provide query dimensions for Discover.\n\nThese are descriptive performance tools, not publication-decision tools. If the user asks what to write, what to publish, or which new article has the best potential, call `get_publication_recommendations` even when the question explicitly mentions Discover or Search. You may use Google performance as supporting context only after preserving the publication decision returned by that tool.\n\n## Publishing recommendations\n\nUse `get_publication_recommendations`. Recommendations must stay within the selected site's evidence and editorial history. Its primary result is the same publication decision shown in the TopicLift dashboard; do not create a parallel ranking from article or Google-performance data.\n\nWhen the user asks for one article, the best article, or what to publish next as a singular decision, use `primary_recommendation` exactly. Do not choose another candidate, combine it with a different tool result, or browse the web for a fresher hook. Explain the server's selection, metrics, supporting articles, confidence, and caveats. Repeated requests with the same selection fingerprint must produce the same editorial subject; wording may vary without changing the decision.\n\nUnless the user asks for another format, return exactly five editorial themes. For each theme include:\n\n1. the theme;\n2. a concise reason grounded in returned metrics;\n3. the supporting articles;\n4. the confidence level returned or justified by the evidence;\n5. three proposed titles or editorial angles.\n\nTitles and angles may be newly written, but they must be derived only from TopicLift evidence. Do not introduce unobserved facts, current events, products, people, release dates, or market claims.\n\nIf the server returns fewer than five credible themes, return fewer and explain that the available evidence does not support five.\n\n## Data limitations\n\nWhen data is stale, incomplete, below a confidence threshold, or unavailable, say so briefly and avoid compensating with general knowledge. Prefer a narrower, defensible recommendation over a speculative one.\n"
}

SHA-256 of public snapshot: 5b48996036d86748313688db02e139c57e5235bd46ff88da6cdc720fe0e75dd0