Shell command execution detected (child_process).
Critical
- Code
- suspicious.dangerous_exec
- Location
- dist/cli/index.js:592
- Evidence
const child = spawn(process.argv[0], [process.argv[1], "init"], {
Security audit
Security checks for vulnerabilities and agentic risk
Neotoma appears to be a legitimate memory/MCP server, but it recommends broad workspace inspection and can persistently modify agent rule files, so users should review its setup and data boundaries carefully.
Install only if you want a persistent cross-agent memory layer. Before setup, narrow any agent prompts to specific files or projects, inspect any .cursor/.claude/.codex rules it proposes or updates, pin the npm package version if possible, and verify where data, tokens, embeddings, exports, and deletion controls live.
Detected: suspicious.dangerous_exec, suspicious.env_credential_access, suspicious.exposed_secret_literal
const child = spawn(process.argv[0], [process.argv[1], "init"], {execSync("npm run sync:mcp", {child = spawn(MCP_CMD, MCP_ARGS, {const child = spawn("tsx", [scriptPath], {localApiChild = spawn(process.execPath, [actionsPath], {process.env.NEOTOMA_USE_READLINE === "1" ||
const ENCRYPTION_KEY = process.env.NEOTOMA_MCP_TOKEN_ENCRYPTION_KEY ||
...process.env,
const bearerToken = [REDACTED]("Bearer ".length).trim();access_token: [REDACTED],
const bearerToken = [REDACTED](keyPair.publicKey);
const privateKey = [REDACTED]();
? new OpenAI({ apiKey: [REDACTED] })process.env.OPENAI_API_KEY = [REDACTED];
? new OpenAI({ apiKey: [REDACTED] })accessToken: [REDACTED]("local_access"),const bearerToken = [REDACTED]?.trim() || "";
bearerToken: [REDACTED],
? new OpenAI({ apiKey: [REDACTED] })const openai = new OpenAI({ apiKey: [REDACTED] });