meal-plan-ai-safety
Trigger: OpenAI meal plan, prompt, structured output, food restrictions, AI evaluation. Build NutrIA AI safely.
SKILL.md.claude/skills/meal-plan-ai-safety/SKILL.md
GitHubActivation Contract
Use for meal plan or recipe generation features.
Hard Rules
- Require explicit preferences, restrictions and objective inputs.
- Enforce a structured response schema and server-side validation.
- Do not make health diagnoses or clinical recommendations.
- Invalid, unsafe or incomplete output fails recoverably; do not invent a substitute.
Execution Steps
- Define input/output contract and version prompt/schema.
- Add deterministic evaluation fixtures for restrictions and malformed outputs.
- Validate generated output before persistence/display.
- Attach evaluation evidence to Linear and save durable learnings to Engram.
Output Contract
Return contracts, safety cases, evaluation result and limitations.