Skip to content

Add Agent Trust Bench -- adversarial payment-flow benchmark (138 profiles, OWASP LLM Top-10) #46

Description

@chopmob-cloud

Suggesting Agent Trust Bench for the Benchmarks & Datasets section.

Agent Trust Bench -- https://agent-trust-bench.algovoi.co.uk

Live adversarial benchmark for AI agents that handle x402/A2A payment flows. 138 profiles across 30 threat categories, mapped to OWASP LLM Top-10 (LLM01 Prompt Injection, LLM06 Sensitive Info Disclosure, LLM07 Insecure Plugin Design, LLM08 Excessive Agency, LLM09 Overreliance, plus FATF financial-regulation scenarios).

What makes it distinct from static datasets: the bench is a live endpoint. Agents make real HTTP calls against adversarial payment challenges (fake-signed, $1 USDC cap enforced server-side) and pass or fail based on their actual decisions -- not pattern matching against a static corpus.

Three reference personas (safe / aggressive / unbounded) for cross-LLM or cross-framework comparison. Machine-readable discovery at:

GET https://agent-trust-bench.algovoi.co.uk/.well-known/x402.json

MIT-licensed runner. No account required. 30-day responsible-disclosure window for profile vulnerabilities.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions