Security audit
Security checks for vulnerabilities and agentic risk
The skill appears to be a real multi-agent communication hub, but it needs Review because some member-level tools are under-scoped and optional autonomous executors can send task content to external LLM or HTTP endpoints.
Install only in a trusted local agent group. Before enabling autonomous execution, decide whether task descriptions, context, and results may be sent to HOST_EXEC_ENDPOINT or third-party LLM providers. Lock down HUB_AUTH_TOKEN/admin access, narrow activation triggers, and consider fixing member-level authorization listing and pipeline/dependency ownership checks before production use.
opts.baseUrl ??
process.env.HOST_LLM_BASE_URL ??
(this.provider === "anthropic"
? "https://api.anthropic.com/v1/messages"
: "https://api.openai.com/v1/chat/completions");
this.apiKey = opts.apiKey ?? process.env.HOST_LLM_API_KEY ?? "";
this.model =process.env.HOST_LLM_BASE_URL ??
(this.provider === "anthropic"
? "https://api.anthropic.com/v1/messages"
: "https://api.openai.com/v1/chat/completions");
this.apiKey = opts.apiKey ?? process.env.HOST_LLM_API_KEY ?? "";
this.model =
opts.model ??if (!reqId) {
throw new Error("request_authorization 未返回 request_id，无法挂起等待");
}
// 若 Hub 直接通过信任窗口自动批准（AUTH_AUTO_APPROVE），立即 resolve
if (res?.status === "approved") {
this.completeAuth(reqId, "approved");
return;expires_at: number;
}
declare const AUTH_REQUEST_TTL_MS: number;
declare const AUTH_AUTO_APPROVE: boolean;
declare class AuthorizationService {
/** 请求敏感操作的授权，返回新建的 pending 请求（含 request_id） */
createRequest(agentId: string, op: CreateAuthOp): AuthRequestRow;private getById;
}
export declare const authorizationService: AuthorizationService;
export { AUTH_REQUEST_TTL_MS, AUTH_AUTO_APPROVE };},
});
// 可选快路径：开启自动批准且存在有效信任窗口 → 直接 approved
if (AUTH_AUTO_APPROVE && this.hasValidGrant(agentId, op.type)) {
return this.resolve(id, "approved", "system", "auto-approve via valid grant");
}
return this.getById(id);});
// 可选快路径：开启自动批准且存在有效信任窗口 → 直接 approved
if (AUTH_AUTO_APPROVE && this.hasValidGrant(agentId, op.type)) {
return this.resolve(id, "approved", "system", "auto-approve via valid grant");
}
return this.getById(id);
}Detected: suspicious.env_credential_access