system-explorer

System Explorer

Single entry point that orchestrates three sub-skills into a complete learning pipeline: discover systems, analyze in depth, and generate an interactive course -- with user checkpoints between each phase.

First-Run Welcome

When triggered, greet the user and explain the pipeline:

I can help you explore and learn any software system -- from discovery to a full interactive course.

Here's how it works:

1. Discover -- Find and compare systems that match what you're looking for
2. Analyze -- Deep-dive research with layered depth (beginner to advanced)
3. Generate -- Build an interactive multi-page HTML course you can browse

Just tell me what you want to explore:

• A specific system -- "Explore Kafka"
• A category -- "I want to understand message queues"
• A requirement -- "Help me find a database for time-series data"

Flow

User query
    |
    v
+---------------------+
|  1. Clarify target   |  Ask questions if ambiguous
+----------+----------+
           v
+---------------------+
|  2. Ask output dir   |  User chooses where artifacts go
+----------+----------+
           v
+---------------------+
|  3. system-finder    |  Discover & compare systems
|     -> comparison    |  User picks system(s)
+----------+----------+
           v  Checkpoint
+---------------------+
|  4. system-analyzer  |  Deep-dive research
|     -> analysis files|  User reviews findings
+----------+----------+
           v  Checkpoint
+---------------------+
|  5. system-to-course |  Generate HTML site
|     -> [output-dir]  |  User reviews output
+----------+----------+
           v
     Done

Smart Shortcuts

Skip phases when they are not needed:

• Specific system named ("explore Kafka") -- Skip finder and go directly to analyzer. Confirm with user: "You want to explore Apache Kafka. Shall I start the deep-dive analysis, or would you prefer to see alternatives first?"
• Existing finder-report.md -- If [output-dir]/finder-report.md exists, offer to skip finder: "I found an existing finder report. Want to use it, or start fresh?"
• Existing analysis -- If [output-dir]/analysis.json or [output-dir]/analysis.md exists, offer to skip to course generation: "There's already an analysis on file. Jump to course generation, or redo the analysis?"
• Multiple systems -- If the user wants to compare multiple systems, run analyzer for each (potentially in parallel using the Agent tool), then generate a combined or separate course site per system.