adversarial_input_stress_tester
ActiveTool of @gapup/mcp-knowledge
declared in 0.2.0
An asynchronous risk assessment tool that evaluates AI model resilience against adversarial inputs following NIST AI Risk Management Framework (RMF) red-teaming protocols. Designed for security and compliance personas, it accepts model outputs or decision boundaries and returns structured risk scores, failure modes, and adversarial examples. Requires async:true to avoid timeout errors. Outputs include status, warnings, and source references.
Parameters schema
{
"type": "object",
"required": [
"modelOutput"
],
"properties": {
"async": {
"type": "boolean",
"description": "If true, returns a job_id immediately (<200ms) instead of waiting for the result. Poll the result with job_result(job_id). Use for slow tools to avoid client timeouts."
},
"maxTests": {
"type": "integer",
"default": 10,
"maximum": 100,
"minimum": 1,
"description": "Maximum number of adversarial tests to run"
},
"modelOutput": {
"type": "string",
"description": "The AI model's output or decision to be stress-tested"
},
"adversarialDataset": {
"type": "array",
"items": {
"type": "string"
},
"description": "Optional custom adversarial inputs to test"
},
"sensitivityThreshold": {
"type": "number",
"default": 0.9,
"maximum": 1,
"minimum": 0,
"description": "Threshold for flagging high-risk adversarial examples"
}
}
}No endpoints wrapped at confidence ≥ 0.70.
Parent server
@gapup/mcp-knowledge
https://github.com/getgapup/gapup-mcp
2/7 registries