Communitygithub.com

Alessiawondering/ielts-learning-studio-skill

Reusable Codex Skill for building a personalized IELTS learning studio from Google Drive materials.

ielts-learning-studio-skill とは?

ielts-learning-studio-skill is a Codex agent skill that reusable Codex Skill for building a personalized IELTS learning studio from Google Drive materials.

対応~Claude CodeCodex CLI~Cursor
npx skills add Alessiawondering/ielts-learning-studio-skill

お気に入りのAIに質問する

このエージェントスキルを事前に読み込んだ状態で新しいチャットを開きます。

ドキュメント

Build IELTS Learning Studio

Create a complete, usable learning site for the current learner. Treat any existing IELTS site as a behavioral reference only. Never copy another learner's Drive URL, documents, dates, topics, saved points, favorites, mistakes, or deployment identity.

Start With The Learner

Gather only missing decisions, in one group of at most three questions:

  1. Ask for the learner's Google Drive folder URL or whether to begin with local uploads/manual paste.
  2. Confirm the learning categories and enabled modes. Default to Listening, Reading, Speaking, Writing, notes, flashcards, Chinese-to-English testing, favorites, and a mistake notebook.
  3. Confirm storage expectations: device-local persistence by default, or authenticated cross-device persistence when explicitly requested.

Proceed immediately when the request already answers these questions. Do not require sample documents before producing a working first screen.

Read product-contract.md before implementation. Read customization-and-privacy.md when configuring Drive, persistence, sharing, or deployment. Use acceptance-tests.md for final verification.

Build Workflow

  1. Inspect the workspace and preserve its framework and design conventions. When .openai/hosting.json exists, use the Sites build and hosting workflow.
  2. Create a first-run setup that stores this learner's folder URL and preferences. Do not hardcode a reference learner's values as defaults.
  3. Implement the ingestion pipeline before polishing secondary UI:
    • list supported files from the learner's Drive folder;
    • parse DOCX yellow highlights and tracked revisions without a count limit;
    • extract worksheet vocabulary from PDFs;
    • extract valuable language from ordinary text when no highlight or correction exists;
    • fingerprint files so additions, edits, and deletions update the site cleanly.
  4. Standardize every extracted point before it reaches notes, cards, or tests:
    • use a reusable lemma or dictionary form;
    • use singular nouns unless plural is grammatically required;
    • preserve passive participles such as be accompanied by;
    • preserve required gerunds such as prevent someone from doing something;
    • keep dependent prepositions and complete collocations together;
    • split a sentence containing multiple learnable words or collocations into separate points;
    • exclude names, formatting debris, and low-value corrections.
  5. Enrich each accepted point with an accurate Chinese meaning, usage, example, pronunciation, collocations, source sentence, date, topic, and one or more IELTS categories. Never show placeholder explanations when enrichment fails; retry or mark the item for review.
  6. Keep one logical knowledge point when the same expression appears in several categories, but display it in every applicable category.
  7. Build the complete learning loop: notebook, study cards, speech playback, favorites, scoped testing, hints, reveal-after-three-errors, persistent mistake notebook, and daily/topic archive.
  8. Persist learner state. Device-local storage must survive refresh and browser exit. Use server persistence only when the user asks for account-based or cross-device behavior.
  9. Validate the behaviors in acceptance-tests.md, build successfully, and publish privately unless the learner explicitly chooses another access model.

Adaptation Rules

  • Allow the learner to change labels, enabled categories, theme naming conventions, test scopes, and visual direction without weakening extraction correctness.
  • Infer date, topic, and category from filenames and document headings. Provide editable fallbacks when inference is ambiguous.
  • Prefer a quiet, work-focused interface suitable for daily repetition. Keep controls compact and responsive; do not turn the product into a marketing landing page.
  • When the learner has no Drive folder yet, keep uploads and pasted text fully functional and make Drive connection an optional next step.
  • Explain the public-link limitation honestly. A public Drive folder MVP is not equivalent to private OAuth access.

Completion Standard

Finish with the learner's independent site URL and a concise summary of enabled capabilities. State any unavailable integration plainly. Do not expose source credentials, internal project IDs, reference learner data, or deployment internals.

関連スキル