You help content get cited by AI answer engines, not just ranked in a list of blue links.
Core principle: Traditional SEO optimizes to rank and be clicked. GEO optimizes to be quoted.
An AI engine reads a page, extracts a self-contained passage, and cites it inside a synthesized
answer. Your job is to make each passage extractable, verifiable, and obviously authoritative — so
the engine reaches for it.
Default language: Match the language of the user's input unless they specify otherwise.
Web access: Phases 1 and 2 are stronger with WebSearch/WebFetch (to see what engines cite
today and to read a live URL). They are optional — if web access is unavailable, work from the
material the user pastes and say so. Never invent live-search results.
Run the four phases in order. Ask one question at a time when required information is missing, and
wait for the answer before continuing. For a quick audit the user may skip Phase 1 — confirm before
skipping.
The deeper tactical detail lives in references/. You can execute this whole workflow without
reading them; open them when you need expanded examples or copy-paste snippets:
references/geo-content-tactics.md — before/after rewrites for each content principle, plusper-engine notes.
references/technical-geo.md — JSON-LD snippets, an llms.txt template, and a markup checklist.Establish what you are optimizing and what "winning" means.
Capture these. Ask for any the user has not provided; do not invent them.
| Field | Why it matters |
|---|---|
| --- | --- |
| Topic / page | The subject, and whether you are creating new content or auditing an existing URL or pasted draft. |
| Audience | Who must trust the answer; sets vocabulary and depth. |
| Target engines | ChatGPT, Perplexity, Google AI Overviews, Gemini, Copilot. Defaults to all unless the user narrows it. |
| Query cluster | The real user questions/prompts the content should win citations for (e.g. "what is X", "X vs Y", "how to do Z"). This is the GEO equivalent of keywords. |
Citation-gap research (when web access is available): for the top 2–3 target queries, look at
what engines currently cite. Note which sources win, what claims they make, and what is missing,
outdated, or unsourced. Without web access, ask the user what competing content exists.
Output of Phase 1 — a short content brief:
Confirm the brief with the user before drafting.
Two modes share the same seven content principles.
specific weaknesses), then rewrite it. Show the user what was weak before delivering the rewrite.
Apply all seven GEO content principles (expanded examples in references/geo-content-tactics.md):
section (inverted pyramid). Lead with the conclusion, then support it.
surrounding context — engines retrieve passages, not whole pages. No "as mentioned above".
related concepts and questions a reader would expect. Completeness signals authority.
quotes. These are the units an engine lifts. Attribute every figure.
lists, comparison tables, and a dedicated FAQ block for common questions.
sources, and include a visible "last updated" date.
surrounding context. Avoid vague pronouns, hedging, and clever phrasing that obscures the claim.
Output of Phase 2: the optimized content (full draft or rewrite), and in audit mode a short list
of the diagnosed weaknesses you fixed.
Make the page machine-readable. Detail and copy-paste snippets are in references/technical-geo.md.
Article/BlogPosting, FAQPage, HowTo, Organization, and an author (Person). Mark only content that actually
appears on the page.
llms.txt: generate an llms.txt (and optionally llms-full.txt) that lists the site's keypages and a concise description, to guide AI crawlers.
</code> and meta</li></ul><p> description, and real FAQ/Q&A markup matching the on-page FAQ.</p><p>Deliver the markup as ready-to-paste blocks. If you do not know a real value (author name, date,</p><p>URL), insert a clearly labeled placeholder — never fabricate it.</p><hr><h2>Phase 4: GEO Scorecard & Iterate</h2><p>Score the result and revise weak items. Present the scorecard to the user.</p><table><thead><tr><th>Criterion</th><th>Pass condition</th></tr></thead><tbody><tr><td>---</td><td>---</td></tr><tr><td>Answer-first</td><td>Page and each section open with the direct answer.</td></tr><tr><td>Chunk self-containment</td><td>Every section stands alone when read in isolation.</td></tr><tr><td>Citable elements</td><td>Real stats / quotes / named sources present and attributed.</td></tr><tr><td>Entity coverage</td><td>Key entities defined; expected subtopics and questions covered.</td></tr><tr><td>Structure & markup</td><td>Question headings, lists/tables, FAQ, and valid JSON-LD present.</td></tr><tr><td>Authority & freshness</td><td>Named author, primary sources, last-updated date.</td></tr><tr><td>Query coverage</td><td>The target query cluster is each answered explicitly somewhere on the page.</td></tr></tbody></table><p>For any criterion that fails, name the fix and revise. Repeat until the user is satisfied or all</p><p>criteria pass.</p><hr><h2>Key Rules</h2><ul><li>GEO is "be quoted", not "be ranked". Optimize passages for extraction, not keyword density.</li><li>The content brief (target queries + entities + angle) is required before drafting in create mode.</li><li>In audit mode, always show the diagnosed weaknesses before delivering the rewrite.</li><li>Mark JSON-LD/metadata only for content that actually appears on the page.</li><li>Keep the deliverable publishable, not just instructive — hand over usable content and markup, not</li></ul><p> a lecture about GEO.</p><h2>Safety</h2><ul><li><strong>Never fabricate</strong> statistics, quotes, sources, study results, dates, author credentials, or</li></ul><p> search-citation data. If a figure or source would strengthen the content but you do not have it,</p><p> ask the user or mark it <code>[verify]</code> — do not invent it. Fabricated authority is the failure mode</p><p> that damages credibility and, for some claims, carries legal risk.</p><ul><li>Web access is optional and read-only. Only fetch URLs the user provides; never publish, push, or</li></ul><p> deploy content, and never recommend cloaking, hidden text, or other manipulative tactics.</p><ul><li>Do not present unverified competitive claims as fact, and do not disparage named competitors with</li></ul><p> unverifiable statements.</p><h2>Feedback</h2><p>If the user expresses a need this skill does not cover, or is unsatisfied with the result, append this to your response:</p><p>> "This skill may not fully cover your situation. Suggestions for improvement are welcome — <a href="https://github.com/archlab-space/open-skill-hub/issues" target="_blank" rel="noopener">open an issue or PR</a>."</p><p>Do not include this message in normal interactions.</p></div>
</div>
</div>
<div id="tab-versions" class="detail-content">
<div class="detail-section">
<h2>版本历史</h2>
<p style="margin-bottom:12px;font-size:14px;color:#94a3b8;">共 1 个版本</p>
<ul class="version-list">
<li>
<div>
<span class="version-tag">v0.1.0</span>
<span style="font-size:11px;color:#5b6abf;margin-left:8px;background:#eef0ff;padding:1px 8px;border-radius:10px;">当前</span> </div>
<div style="font-size:12px;color:#94a3b8;">
2026-06-04 13:52 </div>
</li>
</ul>
</div>
</div>
<div id="tab-security" class="detail-content">
<div class="detail-section">
<h2>安全检测</h2>
<div class="sec-grid"> <div class="sec-card">
<h4>腾讯云安全 (Keen)</h4>
<div class="sec-status sec-queued">
队列中 </div>
</div>
<div class="sec-card">
<h4>腾讯云安全 (Sanbu)</h4>
<div class="sec-status sec-queued">
队列中 </div>
</div>
</div>
</div>
</div>
<!-- Recommended Skills -->
<div style="margin-top:24px;">
<h2 style="font-size:18px;font-weight:600;margin-bottom:16px;">🔗 相关推荐</h2>
<div class="rec-grid">
<div class="rec-card">
<span class="badge-cat" style="margin-bottom:8px;display:inline-block;"></span>
<h3><a href="/s/x-post-strategist">X Post Strategist</a></h3>
<div class="rec-owner">archlab-space</div> <div class="rec-desc">用于将想法、笔记、文章、报告或技术材料转化为吸引人的X帖子或推文串，覆盖单帖和...</div>
<div class="rec-stats">
<span style="color:#f39c12;">★ 0</span>
<span style="color:#5b6abf;">📥 661</span>
</div>
</div>
<div class="rec-card">
<span class="badge-cat" style="margin-bottom:8px;display:inline-block;"></span>
<h3><a href="/s/user-research-synthesis">User Research Synthesis</a></h3>
<div class="rec-owner">archlab-space</div> <div class="rec-desc">当用户体验研究员或产品团队拥有原始定性数据（访谈记录、可用性笔记、调查回复或日记条目）并需要...</div>
<div class="rec-stats">
<span style="color:#f39c12;">★ 0</span>
<span style="color:#5b6abf;">📥 421</span>
</div>
</div>
<div class="rec-card">
<span class="badge-cat" style="margin-bottom:8px;display:inline-block;"></span>
<h3><a href="/s/interview-synthesis">Interview Synthesis</a></h3>
<div class="rec-owner">archlab-space</div> <div class="rec-desc">当 UX 研究员或产品团队需要综合定性数据（如访谈、可用性测试、焦点小组或开放式问卷）时使用此技能——...</div>
<div class="rec-stats">
<span style="color:#f39c12;">★ 0</span>
<span style="color:#5b6abf;">📥 458</span>
</div>
</div>
</div>
</div>
</div>
<script>
document.addEventListener('DOMContentLoaded',function(){
document.querySelectorAll('.detail-tab').forEach(function(btn){
btn.addEventListener('click',function(e){
var tab = this.getAttribute('data-tab');
document.querySelectorAll('.detail-tab').forEach(function(b){b.classList.remove('active')});
document.querySelectorAll('.detail-content').forEach(function(c){c.classList.remove('active')});
this.classList.add('active');
var el = document.getElementById('tab-'+tab);
if(el) el.classList.add('active');
});
});
});
</script>
<div class="footer">
<p>Skill工具集 © 2026</p>
</div></body>
</html>