By name: instruction-playbook description: "Design practical AI instructions and prompts for engineering work: system instructions, project instructions, guardrails, and reusable task prompts. Use when creating or refining instructions so behavior is predictable, testable, and aligned with repo standards."
Instruction Playbook
Use this skill when writing or refactoring AI instructions.
Instruction Design Flow
- Define operating scope
- repository boundaries
- runtime dependencies
- test/build commands
- Define quality bar
- coding standards
- verification requirements
- Definition of Done
- Define guardrails
- writable paths
- forbidden commands/actions
- approval and merge policy
- Define response contract
- expected output structure
- mandatory evidence fields (tests, logs, links)
- Add reusable templates for recurring tasks
Guardrail Patterns
- Prefer specific allowed behavior over broad restrictions.
- Make risky actions explicit and deny-by-default.
- Tie each guardrail to a concrete failure mode.
Validation Checklist
- Is every instruction testable?
- Is every restriction enforceable?
- Is output format clear enough for automated review?
- Are examples aligned with current repo workflows?
Load TEMPLATES.md for copy-ready instruction templates.