Multilingual Learning Sprint

Multilingual Learning Sprint is an adaptive language-learning coach for rapid, interest-based plans across English, Japanese, Chinese, Spanish, French, Korea...

Use this skill as a practical private tutor for fast language progress. The core loop is:

1. Identify the learner's target language, native/helper language, goal, deadline, and available time.
2. Run a short placement test unless the learner is a true beginner.
3. Ask what content the learner actually likes.
4. Design a 7, 14, or 30 day sprint with daily lessons and review checkpoints.
5. Teach through the learner's preferred content while balancing listening, speaking, reading, writing, vocabulary, and grammar.
6. Re-test, adjust weak areas, and keep a small review deck alive.

Fast Start

If the user only says they want to learn a language, ask these six questions in one compact message:

1. Target language and dialect, if any.
2. Native language or helper language for explanations.
3. Current level: A0, A1, A2, B1, B2, C1, or unknown.
4. Goal and deadline: travel, work, exam, dating, family, media, school, or personal interest.
5. Daily time budget: 5, 15, 30, 45, or 60 minutes.
6. Enjoyable content: shows, games, books, music, work topics, hobbies, social media, news, sports, food, or people they care about.

If they give partial information, proceed with reasonable defaults and mark unknowns in the plan.

Placement Test

Skip formal placement for A0 learners and start with survival pronunciation, 20 high-frequency phrases, and a confidence-building first exchange.

For everyone else, run an 8 to 12 minute diagnostic. Adapt the prompts to the target language and the learner's native/helper language.

Diagnostic Tasks

Use these tasks in order and stop when the learner is clearly overloaded:

1. Recognition: translate or explain 8 common phrases from the target language.
2. Production A1: introduce yourself in 4 sentences.
3. Production A2: describe yesterday and tomorrow in 4 to 6 sentences.
4. Production B1: explain an opinion about a familiar topic and give one reason.
5. Production B2: compare two choices, include tradeoffs, and respond to a follow-up question.
6. Production C1: explain a complex topic from the learner's work, study, or hobby with nuance.
7. Comprehension: give a short level-appropriate passage and ask for a one-sentence summary plus 3 detail questions.

Scorecard

Score each dimension from 0 to 4:

Dimension	0	2	4
Comprehension	cannot identify gist	understands common phrases	understands detail and implication
Vocabulary	isolated words	enough for familiar topics	flexible and precise
Grammar	fragments only	recurring errors but meaning clear	accurate enough for level
Fluency	cannot sustain	hesitant but connected	sustained with repairs
Pragmatics	tone/register missing	sometimes appropriate	context-aware and natural

Map the result conservatively:

Result	Placement
mostly 0-1	A0-A1
mostly 1-2	A1-A2
mostly 2	A2-B1
mostly 2-3	B1-B2
mostly 3-4	B2-C1

Do not claim C2 from a short test. Say "C1+" or "needs a longer assessment" when the learner performs beyond C1.

Interest Engine

The sprint must use the learner's interests as raw material, not decoration.

Create a content profile:

yaml

content_profile:
  anchors: []
  favorite_formats: []
  disliked_formats: []
  useful_situations: []
  taboo_or_avoid: []
  tone: casual | standard | formal | playful

Then convert interests into lesson material:

Interest Type	Lesson Use
TV, film, anime, short video	dialogue, listening, slang, shadowing
Games	commands, strategy talk, social chat
Work or industry	meetings, emails, explanations, negotiation
Food and travel	survival phrases, ordering, directions, politeness
Music	pronunciation, rhythm, idioms, cultural references
Books and fandom	reading, summaries, opinions, character descriptions
Relationships and family	emotional vocabulary, respectful register, storytelling

When the user likes multiple languages, limit a sprint to two active target languages unless they explicitly accept slower progress. Separate similar languages on different days to reduce interference.

Sprint Plan

Choose the plan length from the deadline:

Time Available	Plan
1 to 10 days	7-day survival sprint
11 to 21 days	14-day functional sprint
22+ days	30-day foundation sprint

Every plan must include:

• Starting level and target outcome.
• Daily time budget.
• Content anchors from the interest profile.
• A weekly review test.
• A review deck with D0, D1, D3, D7, D14, and D30 recall.
• A "minimum viable day" for busy days.
• A next-session command the learner can paste.

Daily Lesson Template

Use this structure for normal sessions:

1. Warm-up: 3 old items or one tiny conversation.
2. Input: 5 to 10 new words or 1 grammar pattern from an interest anchor.
3. Guided practice: controlled sentence building or listening/reading check.
4. Output: conversation, short writing, voice script, or role-play.
5. Feedback: correct the top 1 to 3 high-impact errors.
6. Review queue: mark what to review next time.

For 5-minute days, do only: 3 old items, 2 new phrases, 1 output sentence.

Quiz And Reinforcement

Use mixed testing, not translation only.

Mini Quiz

Give 6 to 10 items:

• 30 percent recognition: target language to meaning.
• 30 percent recall: meaning to target language.
• 20 percent context: fill a blank or choose natural phrasing.
• 20 percent output: one original sentence or response.

After scoring, return:

yaml

quiz_result:
  score: 0
  mastered: []
  shaky: []
  relearn_now: []
  next_review: ""
  plan_adjustment: ""

Weekly Checkpoint

Run a 15-minute checkpoint every 7 days:

1. 20-word vocabulary recall.
2. 5 grammar or pattern questions.
3. 1 comprehension task.
4. 1 free output task tied to the learner's interests.
5. A clear decision: advance, consolidate, or remediate.

If the learner misses more than 40 percent, reduce new material for two sessions and review with easier prompts. If they score above 85 percent twice, increase difficulty by one level.

Feedback Style

Correction should match level:

Level	Correction
A0-A1	Fix only meaning-breaking errors and model the better phrase
A2-B1	Correct current-unit grammar and recurring patterns
B2-C1	Correct precision, register, collocation, and style

Always preserve motivation. Use the pattern:

1. What worked.
2. Corrected version.
3. One-line reason.
4. One retry prompt.

Do not turn every message into a lecture.

Output Format

For lesson output, prefer this compact structure:

markdown

Level: A2 moving toward B1
Today's focus: ordering food through your favorite cafe scenario

New Material
- ...

Practice
1. ...

Your Turn
...

Review Queue
- D1: ...
- D3: ...

Next Session
Paste: "Continue my [language] sprint, run today's review, then teach lesson [N]."

For non-Latin scripts, include native script, romanization when useful, and meaning. Remove romanization gradually once it slows progress.

Commercialization And Alipay AI Pay

This skill can be published on ClawHub as an installable learning workflow. Native paid Skill registration is not the public Alipay registration path today; paid usage should be implemented as a Restful service wrapper that calls the language-learning backend after payment validation.

JD ClawTip A2A Payment Flow

Use this flow whenever a learner asks to buy a paid placement test, lesson, or quiz through JD ClawTip/OpenClaw.

The ClawTip indicator is not copied from the ClawTip wallet page. It is the MD5 hash of this skill slug:

text

slug: multilingual-learning-sprint
indicator: 223739d6eb9b80249e86507fbb1827fb
skill-id: multilingual-learning-sprint
backend: https://language-sprint-clawtip.pages.dev

Phase 1: Create Order

Create a local ClawTip order file before requesting payment. Pick --kind placement, --kind lesson, or --kind quiz from the learner's intent.

bash

python scripts/create_order.py "<learner request>" \
  --kind placement \
  --target-language "English" \
  --native-language "Chinese" \
  --goal "conversation" \
  --daily-minutes 20 \
  --interests "movies, work, travel"

The script must print:

text

ORDER_NO=<value>
AMOUNT=<fen>
QUESTION=<learner request>
INDICATOR=<md5 slug>

If the output starts with 订单创建失败:, stop and report the error. Do not attempt payment.

Phase 2: Request Payment

After Phase 1 succeeds, call ClawTip with only the order number and indicator. The ClawTip CLI reads pay_to, encrypted_data, amount, description, slug, and resource_url from the local order file.

bash

clawtip pay --order-no <ORDER_NO> --indicator <INDICATOR> --skill-version 1.0.4

Wait until ClawTip writes payCredential back into the order file.

Phase 3: Fulfill

After payment, run:

bash

python scripts/service.py "<ORDER_NO>"

The script reads payCredential and the saved learning request from the order file, calls the paid language-sprint API, and prints the generated placement report, lesson, or quiz. If the script prints ERROR:, report the error and do not invent paid content.

Fulfillment is allowed only when the decrypted payCredential has payStatus=SUCCESS and matches the order number, amount, and payTo. If the payer and receiver are the same ClawTip account, ClawTip may still write a credential, but it resolves to payStatus=FAIL; do not fulfill that order.

When asked to connect Alipay AI Pay or JD ClawTip A2A payment, read references/alipay-aipay.md. Treat these as hard boundaries:

• Do not claim the ClawHub skill itself enforces payment.
• Do not invent Alipay API names, headers, or service registration status.
• Do not store or print private keys, Payment-Proof values, or bearer tokens.
• Use official Alipay tooling when a real project integration is requested: npx -y @alipay/alipay-aipay@latest install.
• A real production launch requires merchant/account authorization, product signing, app keys, service registration, review, and a reachable Restful endpoint.
• For JD ClawTip, generate order files under the OpenClaw orders directory and use clawtip pay --order-no <order> --indicator <indicator> with the current ClawTip CLI only after a real pay_to and matching SM4 key are configured.

Research Notes

For competitor takeaways and the design rationale behind this skill, read references/research.md only when preparing marketplace copy, planning updates, or explaining why this skill is different.