In this article
rai-standards
| Field | Value |
|---|---|
| Kind | skill |
| Source | .github/skills/rai/rai-standards |
| Invocation | Loaded on demand by referencing agents |
| Interactive | No |
What it does
Consolidated Responsible AI standards reference: NIST AI RMF 1.0, AI STRIDE threat-modeling overlay, EU AI Act risk tiers, and an open-standards catalog with phase mapping
When to use it
The RAI Planner loads this standards package for NIST AI RMF mapping, the Microsoft
AI STRIDE overlay, and a paraphrased EU AI Act reference. Use it when an assessment
needs a standards baseline; use the rai-planner reference for phase-specific
workflow guidance instead.
Additional organizational frameworks layer onto the default NIST mapping rather than replacing it. Restricted standards remain citation-only, and regulatory interpretation requires qualified review rather than reliance on a packaged summary.
Example usage
Ask the RAI Planner to load rai-standards for a fictional document-summarization
assistant. Supply its intended users, data categories, oversight design, and a
short organization-authored policy. Request a standards map, not a legal conclusion.
The planner should retain the NIST baseline, map the supplied policy as an attributed extension, and select phase-relevant references for measurement, threat analysis, or impact management. Success is a map from system facts and planned controls to standards references, with missing evidence and regulatory questions visible. Do not paste restricted normative text or treat the output as certification, model evaluation results, or approval to deploy.