agents__creating-folder-agents

You create AGENTS.md files to enforce folder-level AI governance by defining domain context, hard-no constraints, and local coding conventions. This ensures deterministic agent behavior and prevents unsafe generic fixes in critical subsystems.

Constraining AI agent behavior in specific folders is best handled by writing structured AGENTS.md files. These files codify local architecture rules and performance budgets to prevent authoritative but unsafe coding modifications.

Updating existing AGENTS.md files without losing documentation requires augmenting the current content. The skill preserves existing rules while appending new folder constraints and ensuring consistency with repo standards.

You need folder-specific constraints for AI coding agents when local architecture, performance budgets, generated-code boundaries, or legacy semantics make generic fixes unsafe. AGENTS.md files enforce these hard-no rules.

You can use AGENTS.md to define hard-no rules for legacy code subsystems. The files codify legacy semantics and domain context, preventing AI agents from applying generic fixes that violate local invariants.

Limitations of using AGENTS.md for AI workflow governance include relying on the agent's ability to parse and respect structured constraints. It requires consistent updates to maintain alignment with evolving repo standards and local conventions.