prompt-integritylisted

# Prompt Integrity For AI-centric applications, the system prompt is code. This skill validates that any defined AI persona or system prompt is structured safely. ## Responsibilities 1. **Injection Guardrails**: Ensures the generated text contains explicit instructions ignoring malicious user overrides (e.g., "ignore all previous instructions"). 2. **Format Enforcement**: Verifies that the prompt clearly dictates the expected output format (JSON, Markdown, Markdown with Tags) and limits hallucinated fields. 3. **Constraint Verifiability**: Ensures the prompt isn't too ambiguous (e.g., "be helpful and safe") but provides actionable boundaries. ## Checkpoint - If a project uses an AI provider, the `Memory.md` and `anti-hallucination.md` artifacts MUST be checked by this skill. - Outputs actionable suggestions or rewrites for prompt segments that score low on structural density.