agent-guidelines

Agent guidelines enforce safety by standardizing behavior across all tasks, requiring thinking-before-action, exhaustive file scanning, and strict language conventions to ensure consistent and disciplined outputs.

Spec-driven agent guidelines work by loading Markdown operational instructions with YAML frontmatter metadata, enforcing strict input rules and language conventions that govern how agents process and execute tasks across coding and analysis activities.

To stop agents from making assumptions, apply guidelines that enforce a thinking-before-coding phase designed to surface assumptions and uncertainties, while promoting simplicity and avoiding unnecessary features or complexity.

Yes, these agent guidelines apply to every agent activity, including coding, analysis, and messaging, ensuring consistent discipline, exhaustive scanning, and safe operation across all task types.

The best way to prevent missed files is to enforce exhaustive scanning of all files in scope before taking action, a core guideline that ensures agents review everything within the operational scope to prevent misses.

YAML frontmatter is supported for metadata including name and description, and the Markdown body is loaded as operational instructions when activating an agent to enforce the specified safety and discipline rules.