Back to plugin

Security audit

ClawHandoff

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this plugin suspicious because it reports that discovered service details can redirect handoff requests, credentials, and prompt contents to an unapproved endpoint.

Review before installing. Consider whether you will configure credentials or send sensitive prompts, and whether the advertised service endpoints are trusted. The scan reports static data-flow analysis and no dynamic exploit execution, so it identifies a serious risk but does not show that exposure has occurred.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

MCP01

Error
Location
plugin/dist/a2a-client.js:285
Finding

Agent Card-controlled RPC endpoints receive configured credentials and handoff contents ## Remediation Suggestions 1. **Validate discovered endpoints before caching or using them.** - Default to same-origin HTTPS endpoints. - Reject HTTPS-to-HTTP downgrade. - Reject unsupported schemes, embedded URL credentials, and fragments. - Require explicit operator configuration for cross-origin RPC endpoints. 2. **Bind credentials to approved destinations.** - Associate tokens and sensitive headers with exact approved origins. - Do not forward discovery credentials to newly advertised origins automatically. - Configure separate credentials where legitimate cross-origin services require them. 3. **Control redirects.** - Disable automatic redirects for authenticated discovery and RPC requests, or validate every redirect destination before resending sensitive headers or request bodies. 4. **Restrict network reachability.** - Use destination allowlists and egress controls appropriate to the deployment. - Block loopback, link-local, private, and metadata-service destinations unless explicitly required and approved. 5. **Add regression tests.** - Cover malicious `supportedInterfaces`, top-level `url`, and `additionalInterfaces` values. - Verify rejection of cross-origin endpoints and transport downgrade. - Verify that neither bearer tokens nor custom authentication headers reach unauthorized destinations. - Cover discovery and RPC redirects. 6. **Respond to potential prior exposure.** - If an untrusted or compromised Agent Card was used, inspect outbound request records and rotate potentially disclosed credentials.

Content
View full analysis

Vulnerability Details

File Location: plugin/dist/a2a-client.js, lines 183–195, 285–315, 318–336, and 371–386
Vulnerability Type: Credential disclosure and server-side request forgery through unvalidated service discovery
Risk Level: High
Confidence: High; confirmed through static data-flow analysis. No dynamic exploit was executed.

Vulnerable Code

The client constructs authentication headers using configured static headers and an environment-variable credential:

javascript
buildHeaders(protocolVersion, includeJsonContentType = false) {
    const headers = {
        Accept: "application/json",
        ...(this.target.headers ?? {}),
    };
    if (includeJsonContentType) {
        headers["Content-Type"] = "application/json";
    }
    if (protocolVersion === "1.0") {
        headers["A2A-Version"] = "1.0";
    }
    if (this.target.authTokenEnvVar) {
        const token = process.env[this.target.authTokenEnvVar]?.trim();
        if (!token) {
            throw new BridgeError("AUTH_LOGIN_REQUIRED", `Environment variable "${this.target.authTokenEnvVar}" is not set for A2A target "${this.target.targetId}".`, `Export ${this.target.authTokenEnvVar} before running handoffs to "${this.target.targetId}".`);
        }
        if (this.target.authHeaderName) {
            headers[this.target.authHeaderName] = token;
        }
        else {
            headers.Authorization = /^bearer\s+/i.test(token) ? token : `Bearer ${token}`;
        }
    }
    return headers;
}

For A2A v1.0, the Agent Card interface URL becomes the RPC destination without origin or HTTPS enforcement:

javascript
return {
    name,
    version,
    protocolVersion: resolvedProto,
    rpcUrl: new URL(ifaceUrl, `${this.baseUrl}/`).toString(),
    streaming,
};

The corresponding A2A v0.3 selection also accepts a card-supplied destination:

javascript
let rpcUrl = readString(card, "url") ?? this.baseUrl;
const preferredTransport = (readString(card, "pref
...[truncated 5328 chars]

SkillSpector findings are pending for this release.

Endor dependency reachability

Checked Nov 10, 2026 · 12:41 AM UTC

No reachable vulnerabilities reported

Endor dependency scan completed.