A2A Agent Card Security Scanner: Audit Identity, Auth & SSRF

Automated cloud DevSecOps security linter and compliance auditor for the Linux Foundation LF AI & Data / Google Cloud Agent2Agent (A2A) Protocol. Inspects public and staging /.well-known/agent-card.json endpoints to identify cryptographic spoofing, internal SSRF exposure, unauthenticated dangerous primitives, and prompt injection attacks in seconds.

---

⚡ What is the A2A Agent Card Security Scanner?

The A2A Agent Card Security Scanner provides zero-friction, automated security surveillance for autonomous AI agent networks. As enterprise multi-agent meshes scale across corporate boundaries, an agent's digital identity and access contract are governed exclusively by its Agent Card (/.well-known/agent-card.json).

This Actor continuously audits Agent Cards against the normative security requirements defined in A2A Specification §8 & §13, producing normalized Trust Scores (0–100), letter grades (A+ through F), and structured compliance findings for DevSecOps, compliance teams, and agent orchestrators.

---

💰 Transparent Pay-Per-Event (PPE) Pricing

Action / Event	Cost	Details
Actor Initialization (apify-actor-start)	$0.03	Charged once per run container startup.
Card Audited (card-audited)	$0.05	Billed only upon successful audit of a live Agent Card.
Dataset Item Export (apify-default-dataset-item)	$0.002	Automatic per-row dataset export fee.
Unreachable / 404 Targets	$0.00	Strict zero-charge guarantee for failed network fetches.

---

🛡️ The 10 Audited Security Threat Vectors

Rule ID	Severity	Category	What is Detected
A2A-SEC-001	CRITICAL	AGENT_SPOOFING	Unsigned Agent Card: Missing cryptographic JWS signatures in signatures. Violates RFC 7515.
A2A-SEC-002	CRITICAL	CRYPTO_INTEGRITY	Insecure Signature: Algorithm is none or symmetric HS256, or jku origin does not match provider domain.
A2A-SEC-003	HIGH	CLEARTEXT_TRANSPORT	Insecure HTTP: Any endpoint in supportedInterfaces, provider, or auth metadata uses http:// instead of https://.
A2A-SEC-004	CRITICAL	SSRF_EXPOSURE	Internal SSRF Exposure: Interface or webhook points to localhost, RFC 1918 private IPs, or cloud metadata (169.254.169.254).
A2A-SEC-005	HIGH	PROMPT_INJECTION	Adversarial Directives: System prompt override phrases or zero-width Unicode obfuscation detected in metadata or skill examples.
A2A-SEC-006	HIGH	DANGEROUS_PRIMITIVES	Unauthenticated Dangerous Skills: High-privilege actions (shell, sql, payments) lack securityRequirements.
A2A-SEC-007	MEDIUM	DEPRECATED_AUTH	Deprecated OAuth Flows: Declares implicit or password (ROPC) flows, officially deprecated in a2a.proto.
A2A-SEC-008	MEDIUM	CREDENTIAL_LEAK	Query Param API Key: API key location is "query", exposing secrets in web server logs and HTTP referrers.
A2A-SEC-009	HIGH	CAPABILITY_ESCALATION	Unauthenticated Extended Card: Declares extendedAgentCard: true without required authentication schemes.
A2A-SEC-010	LOW	SPEC_COMPLIANCE	Schema Compliance: Missing REQUIRED Protobuf fields (name, description, version, supportedInterfaces, skills).

---

⚖️ Competitor Benchmark

Feature	A2A Security Scanner (This Actor)	Cisco a2a-scanner	Manual Security Review
Deployment Model	1-Click Apify Cloud / Scheduled Cron	Self-hosted Python 3.11+ / UV	Manual penetration testing
Audit Speed	< 3 seconds per card	20–45 seconds	Hours to days
LLM Token Cost	$0.00 (Zero-Token AST)	High ($0.05–$0.20 per scan)	High consulting fees
Automated Alerts	Slack, Webhooks, PagerDuty	Requires custom integration	Email reports
SARIF Export	Available via companion CLI	Custom JSON only	PDF reports

---

💻 Integration Examples

Python (apify-client)

from apify_client import ApifyClient

client = ApifyClient('YOUR_APIFY_TOKEN')

run_input = {
    "agentCardUrls": [
        "https://agent.example.com/.well-known/agent-card.json"
    ],
    "strictMode": True,
    "failOn": "HIGH"
}

run = client.actor("neon_innovation_lab/a2a-agent-card-security-scanner").call(run_input=run_input)

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(f"Agent: {item['agentName']} | Score: {item['trustScore']}/100 | Grade: {item['grade']} | Verdict: {item['verdict']}")
    for finding in item["findings"]:
        print(f"  [{finding['severity']}] {finding['ruleName']}: {finding['message']}")

Node.js (apify-client)

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_APIFY_TOKEN' });

const run = await client.actor('neon_innovation_lab/a2a-agent-card-security-scanner').call({
  agentCardUrls: ['https://agent.example.com/.well-known/agent-card.json'],
  strictMode: false,
  failOn: 'HIGH'
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
items.forEach(report => {
  console.log(`${report.agentName}: Grade ${report.grade} (${report.trustScore}/100) - ${report.verdict}`);
});

More from Neon Innovation Lab

• MCP Server Security Scanner: audit MCP servers for SSRF, prompt injection and unsafe file access.
• Zero-Trust MCP Gateway: redact secrets and PII and block risky calls in front of an MCP server.
• DevOps & Cloud Code Auditor: find cloud cost leaks, table-locking migrations and OWASP gaps.
• Google Maps Leads Scraper: businesses from Google Maps with phones, websites and emails found on their sites.