What problem does it solve?
Writing comprehensive documentation for an entire codebase is slow and inconsistent when done manually, and single-pass AI generation often misses architecture, theory, API, and usage details. This Skill analyzes a project, dynamically decomposes documentation work into dependent tasks, and executes them in parallel waves of agents that share discoveries.
Core Features & Use Cases
- Dynamic Task Decomposition: Scans project type, scale, languages, and modules to generate 5-25 documentation tasks (overview, architecture, implementation, theory, API, usage, synthesis) with dependencies.
- Wave-Based Parallel Execution: Computes execution waves via topological sort (Kahn's algorithm) and spawns concurrent agents per wave with configurable concurrency limits.
- Cross-Agent Context Sharing: Propagates findings through wave summaries, a shared discoveries.ndjson board, and per-task prev_context injection so later documents build on earlier results.
- Use Case: Point it at a scientific computing codebase to produce a full docs suite including LaTeX-formatted theoretical foundations, architecture diagrams, API references, and a navigable index.
Quick Start
Ask the AI to run the project-documentation-workflow skill to generate complete documentation for your project directory, optionally with -y for auto mode and -c to set agent concurrency.
