AI-Shifu Course Creator

Problem Addressed

Creating, editing, optimizing, deploying, and managing courses on the AI-Shifu platform often involves complex workflows and multiple dependencies. Manually navigating extensive documentation or writing verbose instructions can easily lead to missed critical steps, such as authentication, target resolution, syncing, and conflict handling.

Core Mechanism & Key Steps

This skill uses a Task Router mapping to translate user intents into a minimal yet complete sequence of reference files.
- Startup Sequence: On first invocation, it reads language-policy.md and session-controls.md to establish the target language and session control rules.
- Intent Routing: Based on the request type (e.g., creating a new course, partial revision, local-only generation, deployment, or data analysis), it sequentially loads the corresponding references/ files (e.g., authentication.md → course-target.md → authoring-mode.md).
- Workflow Execution: It executes stages in order, covering Segmentation, Orchestration, Generation, Optimization, and Deployment.
- State Synchronization: For platform-bound operations, it strictly enforces pull-before-editing and push-existing-course-content, resolving issues via conflict-convergence.

Boundaries & Caveats

• Analytics routing (analytics/workflow.md) is reserved exclusively for observed metrics of an existing live course (e.g., learner progress, revenue, stuck lessons). Design questions (e.g., "how many lessons should this material become?") remain authoring tasks.
• In platform-bound authoring, no outline or lesson content should be generated until the target is resolved.
• If the target type switches between new and existing, the current route must be aborted and reclassified to prevent entering incompatible stages.
• By default, full-course authoring proceeds through deployment and publication; explicit local/artifact-only paths are selected only upon user request.