Agent Security Advisor

Interactive, agentic tools for AI-agent security — explore the OWASP Agentic / Skills / MCP Top 10, MAESTRO, AIVSS and MITRE ATLAS, score a vulnerability with the canonical AIVSS calculator, and generate a tailored threat model or checklist. Instant and free.

goal hijackexcessive agencysupply chainmemory poisoningcommand injectioninter-agent

Search a concern, or pick a framework to list its controls.

CVSS base scoreThreat multiplier (ThM)Mitigation factor

Rate each of the 10 Agentic AI Risk Factors 0 (absent) → 1 (maximal):

AARF-1 · Autonomy of Action0.0

How much the agent acts without human approval.

AARF-2 · Goal-Driven Planning0.0

Ability to plan multi-step actions toward a goal.

AARF-3 · Self-Modification0.0

Ability to change its own prompts, memory, or behavior.

AARF-4 · Dynamic Tool Use0.0

Breadth/sensitivity of tools it can invoke at runtime.

AARF-5 · Persistent Memory0.0

Long-term memory that carries state across sessions.

AARF-6 · Contextual Awareness0.0

How much external context shapes its decisions.

AARF-7 · Dynamic Identity0.0

Credentials/identity it holds and can assume.

AARF-8 · Multi-Agent Interactions0.0

Coordination with other agents (A2A).

AARF-9 · Non-Determinism0.0

Variability of outputs/actions for the same input.

AARF-10 · Opacity & Reflexivity0.0

How hard its reasoning is to inspect/predict.

—/10

—

Canonical AIVSS v0.8: AIVSS = (CVSS_Base + AARS) × Mitigation_Factor, AARS = (10 − CVSS_Base) × (Factor_Sum / 10) × ThM. Advisory, not an official score.

Generate a ready-to-customize starter document.

Get a grounded, cited answer from the AgentReady corpus (Vertex AI Search) plus the matching framework controls — always live.