documentation-writer-agent

Implementation-aligned documentation is produced by grounding outputs in current project evidence and references. This ensures specs, architecture packs, and runbooks faithfully reflect actual implementation rather than speculated system state.

Writing runbooks and specs without inventing system state requires marking assumptions and gaps clearly while using references to ground outputs. This keeps next steps explicit and ensures outputs reflect actual implementation evidence.

Yes, you can create architecture packs and concise summaries for your project. The process applies across multiple documentation types, ensuring all outputs satisfy frontmatter-driven discovery requirements and structured content sections.

Yes, the documentation writer supports frontmatter-driven discovery requirements by including name and description metadata. It structures content sections and optional references or assets as directed by SKILL.md guidelines.

To mark assumptions and gaps clearly in architecture documentation, the process explicitly identifies missing evidence and highlights next steps. This prevents invented system state and maintains fidelity to the actual implementation.

You should not use an automated documentation agent for writing specs when current project evidence is missing or incomplete. Without evidence or references to ground the output, the documentation risks speculating system state rather than reflecting actual implementation.