The human-authored scope: theory-first, mid-level depth, Python, HTML chapter subpages.
run-playbook dispatches agents in waves and tracks state on a shared to-do board.
Creates the Python venv and installs agent-framework so the real API can be introspected and examples can run.
Wave 0 — Design & scaffold
Turns the brief into toc.yml + outline.
Scaffolds the HTML shell + navigation.
Waves 1…N — per chapter group
Introspects the real 1.9.0 API — no hallucinated names.
Gathers concepts from Microsoft docs.
Fills the HTML slots and writes runnable example code.
Runs every example in the venv — must exit 0 with no API keys.
Verdict: ACCEPT or REVISE with concrete fixes.
Each completed wave is committed, keeping a clean, auditable history.
Cross-chapter check of navigation, terminology, progression, and API consistency.
10 chapters of verified content served as an interactive HTML site. Open it →
The actors (primitives)
🧭 Orchestrator prompt
The driver that runs the whole fleet in waves and tracks progress.
🏛️ Architect agent
Designs the table of contents and chapter outline from the brief.
📚 Theory researcher agent
Explains concepts grounded in Microsoft documentation.
🔍 Library explorer agent
Introspects the installed MAF package for the exact, real API surface.
✍️ Chapter author agent
Writes chapter prose and example code into the page slots.
✅ Code verifier agent
Runs every example until it provably exits clean — no keys required.
🧐 Chapter reviewer agent
Critiques for accuracy and consistency; routes fixes back to the author.
🎨 Frontend builder agent
Scaffolds and finalizes the HTML shell, navigation, and styling.
⚙️ Environment setup skill
Reproducible recipe to build the venv and install the MAF packages.
🧱 Content & code rules instructions
Auto-applied conventions every chapter and example must follow.