feat(airt): model-extraction & membership-inference attack generators

capabilities/ai-red-teaming/capability.yaml

@@ -1,16 +1,18 @@
 schema: 1
 name: ai-red-teaming
-version: "1.9.0"
+version: "1.10.0"
 description: >
   Probe the security and safety of AI applications, agents, and foundation models.
   Orchestrates adversarial attack workflows to discover vulnerabilities in LLMs,
   agentic systems, MCP servers, multi-agent architectures, RAG pipelines, IDE/code
   agents, and custom AI endpoints before they are exploited. Covers jailbreaking,
   prompt injection, data exfiltration, tool manipulation, reasoning attacks, guardrail
   bypass, and more — mapped to OWASP LLM Top 10, OWASP ASI01-ASI10, MITRE ATLAS,
-  and NIST AI RMF compliance frameworks. 45 attack algorithms (41 LLM + 4 adversarial
-  ML samplers), 500+ transforms, an extensive scorer catalog, and 260 bundled harm
-  goals across 25 sub-categories in safety, security, and agentic tiers.
+  and NIST AI RMF compliance frameworks. Also probes traditional black-box ML
+  classifiers: model extraction (model stealing), membership inference (training-data
+  leakage), and model evasion (adversarial examples) across tabular, image, and text.
+  500+ transforms, an extensive scorer catalog, and 260 bundled harm goals across 25
+  sub-categories in safety, security, and agentic tiers.
 
 agents:
   - agents/