What problem does it solve? AI agent instructions often fail because they are bloated, overfit to one-off failures, or written for human reviewers instead of the agent executing them. This Skill provides a disciplined methodology for writing, editing, and reviewing instruction files so agents behave predictably at runtime. ## Core Features & Use Cases - Instruction Writing Guidelines: Apply principles like behavioral WHY, progressive disclosure, and examples-over-explanations to produce instructions agents actually follow. - Edit Protocol: Distinguish minor in-place edits from paradigm shifts, generating _v2 sibling files for structural changes so users can diff and review before adopting. - Skill and Persona Authoring: Enforce the skill_file.md schema with valid YAML frontmatter (name, description, trigger) so files are discoverable in the dynamic catalog. - Use Case: When an agent repeatedly misinterprets a rule, use this Skill to diagnose whether the instruction lacks behavioral reasoning, then rewrite it with a leading WHY pattern instead of stacking rigid ALWAYS/NEVER constraints. ## Quick Start Ask the agent to review an existing skill file and rewrite it following the instructions-architect principles, cutting design rationale and adding behavioral reasoning to each rule.
