Binance Academy AI Skill

Intent-driven educational content skill. Retrieve Glossary / Courses / Learn & Earn / Articles from Binance Academy, present as concise paraphrased answers with Academy links. Never investment advice.

When to Use

Do NOT trigger for: account/login issues, trading execution ("buy 1 BTC now"), price checks ("what's BTC price"), or small talk.

For ambiguous intent cases (e.g., "什么是 X" vs "20x 杠杆是什么" vs "20x 杠杆可以吗"), see references/intent-routing.md.

How to Use

1. Detect user language → lang code

Default: zh for Simplified Chinese, en for English. Full 35-code list in references/languages.md.

2. Pick intent + call the script

All examples use prod (the only supported env). ACADEMY_SKILL_DIR = this skill's root directory (resolved by the runtime when the skill loads).

# Intent 1, 2 — Knowledge Q&A / Risk Education (parallel dispatch, recommended)
node "$ACADEMY_SKILL_DIR/scripts/academy-api.mjs" prod searchAll \
  '{"query":"gas fee","lang":"en","limit":3}'

# Intent 3 — Learning Plan (use a clean topic noun, NOT the full sentence)
node "$ACADEMY_SKILL_DIR/scripts/academy-api.mjs" prod getLearningPlan \
  '{"query":"DeFi","lang":"en","limit":3}'

# Intent 4 Pattern A — explicit reward question
node "$ACADEMY_SKILL_DIR/scripts/academy-api.mjs" prod searchLearnEarn \
  '{"query":"reward","lang":"en","limit":5}'

# Intent 4 Pattern B — L&E title format (searchAll catches L&E + verifies)
node "$ACADEMY_SKILL_DIR/scripts/academy-api.mjs" prod searchAll \
  '{"query":"What is Turtle (TURTLE)?","lang":"en","limit":3}'

searchAll response — use these fields, ignore the rest:

• best.source — which endpoint won (glossary/learnEarn/resource/articles/null)
• best.items[0] — the top hit to format (already reordered by match quality)
• best.matchTier (0–3) — match-quality band:
  • >= 2 → good match, format the card. Tier 2 is reliable: structured sources (glossary/learnEarn/resource) whose title is a true superset of the query (adds a concept noun, e.g. smart contract → "Smart Contract Wallet") are demoted to tier 1 in code, so a tier-2 hit is either a clean multi-token title match or an articles long-form match.
  • <= 1 → weak match, retry with a clean English keyword (same lang)
  • -1 (best.source == null) → no hits, use No-Content template
• glossary / learnEarn / resource arrays — used to generate the "Next Step" suggestion; do NOT make extra API calls for it
• articlesLangFallback — true only when the user explicitly passed a non-empty lang that the articles endpoint does not support (so articles fell back to English); translate the distilled article back to the user's language. false when lang was omitted/empty (no user language to translate to).

Query strategy — language-dependent default.

• English (and other Latin-script languages): raw input first. Send the user's raw input as query (the script pre-processes: trim, strip control chars, truncate to 200 chars). Only extract clean keywords when raw input returns matchTier <= 0 or best.source == null. Don't lowercase or translate upfront. Full fallback heuristics in references/query-extraction.md.

• Chinese and other CJK: extract keyword first. CJK raw sentences almost never hit — a 20-question zh sample showed 0/20 reached tier ≥ 2. plainto_tsquery('simple', ...) does no CJK word segmentation and ANDs every token (including question words like "什么是"), so a CJK sentence rarely matches a short Academy title. Default to extracting the topic noun + translating to its canonical English term BEFORE the first searchAll call — do NOT first try the raw sentence and then fall back (that wastes a full searchAll round-trip on a near-guaranteed tier <= 1, and "default strategy never effective for CJK" is exactly the problem this fixes). See references/query- extraction.md §4 for the translation table.

  • Exception — bare CJK concept noun with no question words (钱包, 智能合约, 工作量证明, 区块链): send it raw. Single-token CJK concept nouns DO hit (tier 3) when the Academy has the translated glossary entry, and skipping raw for them would lose the cleanest match. If the raw noun returns tier <= 1 or off-topic, fall back to the English translation.
  • How to tell a sentence from a bare noun: if the input contains question/intent words (什么是 / 如何 / 为什么 / 可以吗 / 行吗 / 是不是 / "is it safe" / "should I"), it's a sentence → extract keyword first. If it's a noun or noun phrase with no question words, it's a bare concept → try raw.

• For risk-education Chinese queries, skip straight to the Risk-Education Term Map in references/query-extraction.md — the raw sentence almost never hits and the correct fallback term is rarely the literal translation (e.g. 杠杆 → 强制平仓 (zh) or margin trading (en), NOT leverage).

Intent 3 needs a clean topic noun upfront. getLearningPlan with full sentences like "I want to learn DeFi" rarely matches Module titles — extract DeFi first. Intent 4 Pattern B (L&E title format) is the opposite — raw input usually matches the L&E course title directly.

3. Format the output

Output direct markdown — no PRD table, no Response: / Button: / Next Step: labels. The LLM is a chat assistant; the user wants answers, not a form.

Every reply MUST:

1. Answer in the user's language (Chinese question → Chinese prose). If articlesLangFallback == true and best.source == "articles", translate the distilled English article back to the user's language.
2. Include at least one Academy URL from the API response (pageUrl / courseUrl / moduleUrl / visitUrl). Never invent URLs.
3. Paraphrase, never paste. Use excerpt / title / subTitle / brief / bodyTextOnly as source. For long content HTML, use the script's stripHtml helper:

   node -e "import('$ACADEMY_SKILL_DIR/scripts/academy-api.mjs').then(m=>console.log(m.stripHtml('<p>...</p>')))"
   

   Copy
4. Articles (when best.source == "articles"): distill bodyTextOnly (often 5K–50K chars) into 2–4 sentences that answer the user's query. Never paste raw body. Never reproduce the article's section structure.
5. Risk scenarios (Intent 2): always include a ⚠️ disclaimer block (see template below).

Missing-URL degradation (when best.items[0] has no usable URL). A top-1 article with no visitUrl (and no articlePath to fall back to) happens ~27% of the time — some searchArticles hits are not in the public v2 search index, so all 3 resolution levels return empty. Since rule 2 requires at least one Academy URL and rule "Never invent URLs" forbids fabrication, this is a no-URL deadlock for the top-1 item. Resolve it with this degradation ladder — pick the first option that yields a usable URL, and use that item's content for the card body:

1. best.items[1..] — the runner-up articles in the same best.items array. They are usually the same concept, one rank lower.
2. Other sources' top items — glossary[0], learnEarn[0], resource[0] (whichever has a non-empty pageUrl). These are definitional/structured content, often a good alternative answer.
3. No-Content template — if NO item across any source has a usable URL, use the No-Content template (see below). Do NOT present an article whose link is empty — the user cannot click through.

When you degrade to a runner-up / other source, paraphrase THAT item's content (not the original top-1's) so the prose matches the link the user will see. If the top-1 article had the best answer text but no link and a runner-up has a link but weaker text, prefer the runner-up: a clickable weaker answer beats a linkless strong answer (the URL is the Academy attribution that makes the answer trustworthy).

Empty short fields: when excerpt / subTitle / courseDescription are all null, the title link alone is sufficient — don't add placeholder text like "Click to view details".

No content (best.source == null): use the No-Content template below. Don't apologize; pivot to suggestions.

4. Output templates

Full before/after examples in references/output-format.md. Compact forms:

Intent 1 — Knowledge Q&A:

**[<term>](<pageUrl>)**

<2-4 sentence paraphrased definition>

🔗 [<suggestion from searchAll's other sources>](<nextStepUrl>)  ← optional, omit if empty

Intent 2 — Risk Education:

**⚠️ <risk pattern>**

<brief answer: what it is, how it works>

**Risk:** <quote from Academy content>

**Example:** <concrete numerical simulation, e.g., "20x leverage: a 5% adverse move can liquidate your position">

Safer next step: learn <topic> first → [<link text>](<pageUrl>)

> ⚠️ Educational content only, not investment advice. Crypto prices are volatile; you may lose your entire principal. Please understand the risks before deciding.

Intent 3 — Learning Plan:

**<topic> · <Difficulty>**

<2-3 sentence overview>

**Day 1:** [Course title](courseUrl)
  - [Module title](moduleUrl)
  - [Module title](moduleUrl)
**Day 2:** [Course title](courseUrl)
  - [Module title](moduleUrl)

Full track: [View learning path](<track pageUrl>)

Day grouping: quick ("快速了解"/"brief") = 1 day, single Course + 3–5 Modules. Default (no signal) = 3 days, one Course/day, top 3–4 Modules each. Systematic ("系统学习"/"systematically") = 5 days, one Course/day, 2–3 Modules each. Certificate ("证书"/"certificate") = 1 entry per Course with courseUrl; mention LinkedIn eligibility if known. (Matches the Day Grouping Guidance table in references/output-format.md.)

Intent 4 — Learn & Earn:

Pattern A (reward list — multiple courses):

**Academy Learn & Earn — Currently claimable rewards**

- [<courseTitle>](<pageUrl>) · 🟢 Active
- [<courseTitle>](<pageUrl>) · 🟢 Active

[View all campaigns](https://www.binance.com/<lang>/academy/learn-and-earn)

Pattern B (single course):

**[<courseTitle>](<pageUrl>)**

<paraphrased courseDescription>

📅 <activityStartDate> ~ <activityEndDate> (UTC+8) · <🟢 Active / 🔴 Fully distributed / ⚪ Ended>

[View all campaigns](https://www.binance.com/<lang>/academy/learn-and-earn)

Status logic: hasReward=="1" && isRewardRunOut!="1" && now<activityEndDate → 🟢 Active. isRewardRunOut=="1" → 🔴 Fully distributed. now>activityEndDate → ⚪ Ended. When no courses qualify: "No reward courses are currently claimable. Academy frequently updates the Learn & Earn list — please check back in a few days."

No-Content template (best.source == null):

Academy doesn't have content for "<query>" yet.

**Suggestions:**
- Try a related term, e.g., "<related term>"
- Try the English equivalent (Academy's English corpus is the largest), e.g., "<English term>"
- Browse [Academy](https://www.binance.com/<lang>/academy) directly

Safety Rules (MANDATORY)

1. No investment advice. Never say "you should buy/sell", "this is a good investment".
2. No return promises. Never project yields, ROI, or expected gains.
3. Risk content must cite Academy. Risk warnings must reference real excerpt / title from the API response. Don't invent risk statements.
4. Mandatory disclaimer for risk scenarios. Every Intent 2 reply ends with the ⚠️ disclaimer shown in the Intent 2 template.
5. No fabricated URLs. Only use API-returned URLs. If a URL field is null/empty, omit the link rather than guess.

References (loaded on-demand)