Generate inline comments
The README references an authentication flow that was refactored in Q2. The API docs show parameters that were deprecated two releases ago. Function comments describe logic that has been rewritten three times since the comment was added. Developers learn to ignore documentation entirely because reading outdated docs is worse than reading no docs at all. It actively misleads.
How the Code Documentation Writer works
Point the agent at a repository, module, or specific file. It reads the source code and generates documentation matched to the level of abstraction you need. At the function level, it writes docstrings that describe parameters, return values, side effects, and edge case behavior. At the module level, it produces summary descriptions that explain what the module does and how it relates to neighboring modules. At the repository level, it generates README content covering setup instructions, architecture overview, and contribution guidelines.
The agent does not write aspirational documentation about what the code should do. It analyzes what the code actually does: control flow, error handling paths, external dependencies, and data transformations. When the code behavior contradicts existing comments, the agent flags the discrepancy.
Why you need the Code Documentation Writer
Highest value:
- Teams with 50,000 or more lines of code where maintaining documentation manually is a full time job nobody has time for
- Open source projects where contributor onboarding depends on accurate, current documentation and maintainers cannot dedicate resources to writing guides
- Regulated industries (healthcare, finance, defense) where code documentation is a compliance requirement and gaps create audit risk
Lower priority:
- Solo developer projects or small scripts where the code is simple enough to serve as its own documentation
- Throwaway prototypes and experimental branches that will not survive to production
How the Code Documentation Writer compares
The Code Documentation Writer explains what existing code does. The Pull Request Reviewer evaluates whether new or changed code should be merged. Documentation generation is retrospective: it looks at code that exists and describes it. Code review is prospective: it looks at code being proposed and judges it. Teams with documentation debt benefit from the writer first, then pair it with the reviewer to keep new code documented from the start.
