AI-Shifu Skills

中文 README

A unified AI-Shifu skill for MarkdownFlow course production and deployment.

Included Skills

• ai-shifu-course-creator: convert raw course material into optimized MarkdownFlow teaching scripts and deploy them as live AI-Shifu courses through a five-phase pipeline (segmentation, orchestration, generation, optimization, deployment).

The skill includes runnable examples under skills/ai-shifu-course-creator/examples/.

Repository Layout

skills/
  ai-shifu-course-creator/

Usage

The skill keeps SKILL.md as the behavior source of truth. Core skill metadata lives in skills/ai-shifu-course-creator/skill.yaml.

Course Authoring & Deployment Paths

Choose one path based on control needs:

Path A: End-to-End (Recommended)

Use when you want the fastest route from raw material to a live deployed course.

1. Prepare source material (transcript or course documents).
2. Run Phase 1–4 to produce optimized MarkdownFlow lesson scripts.
3. Run Phase 5 to build, import, and publish to the AI-Shifu platform.

Expected artifacts:

• Structured segmentation
• Lesson-by-lesson MarkdownFlow scripts
• Course index and global variable table
• Optimized lesson prompts and risk report
• Live course on the AI-Shifu platform

Path B: Author Only

Use when you need optimized MDF scripts without deploying. Sub-paths:

• Segment only: Phase 1 for semantic segments and manual review.
• Generate only: Phase 3 on pre-existing segments.
• Optimize only: Phase 4 to audit and improve existing scripts.

Path C: Deploy Only

Use when you have pre-existing MDF files ready to deploy:

1. Organize MDF files in a course directory.
2. Run build --course-dir ./course-a/ to generate the import file.
3. Run import --new --json-file ./course-a/shifu-import.json to create the course.
4. Run publish <shifu_bid> to make it live.

Path D: Manage Existing

Use management commands (list, show, update, rename, reorder, delete, publish, archive) on courses already on the platform.

Validate Metadata

python3 scripts/validate_skill_quality.py

Language Policy

Skills are language-flexible for course generation. From a user perspective:

• If you explicitly request an output language, that language is used.
• If you provide target_language, it is used when no stronger explicit instruction exists.
• If neither is provided, the system uses session preference and prompt language signals.
• If language is still ambiguous, it falls back to en-US.
• If you need bilingual output, set bilingual_output: true.

Recommended controls for predictable language output:

• target_language (for example zh-CN, fr-FR, ja-JP)
• bilingual_output (true|false)
• term_policy (preserve|translate|mixed)
• quote_policy (translate_only|original_plus_translation)

AI-Shifu

This suite is part of AI-Shifu's course authoring workflow: https://ai-shifu.com