agent-rules

Generate AGENTS.md files by running a bash-based generator that extracts repository facts, CI rules, and commands to create standardized, machine-readable agent instructions. The generator uses jq and git to analyze the repository and produce root or scoped AGENTS.md files from templates.

AGENTS.md serves as a standardized instruction file for AI agents, replacing scattered onboarding documentation with a single source of truth. It provides machine-readable commands, file maps, and CI rules that agents and maintainers rely on for consistent repository interaction.

CLAUDE.md and GEMINI.md symlinks can be automatically created alongside AGENTS.md to ensure compatibility across multiple AI agent tools. The generation process handles symlink creation, adapting the output for different agent platforms while maintaining a single instruction source.

Bash, jq, and git are required dependencies to verify AGENTS.md commands, as the verification workflow relies on these tools to extract and test documented commands. The scripts enforce verification checkpoints to ensure all documented commands pass before marking items as verified.

Standardize agent documentation across multiple scopes by using detection scripts that identify repository boundaries and generate scoped AGENTS.md files idempotently. The generator applies heuristics and freshness checks to drive consistent updates across all repository levels.

AI agent onboarding fails with inconsistent documentation because agents encounter conflicting instructions across scattered files. Replacing these with a reproducible generator and verification workflow ensures agents interact with up-to-date, machine-readable AGENTS.md files that pass automated verification checkpoints.