autopilot-ai@2.2.0
autopilot (formerly kimiflare): a terminal coding agent for any model with your own key: OpenRouter, Anthropic, OpenAI, Gemini, Bedrock, Vertex, Ollama and more. Per-turn cost, web search, GitHub, headless browser, LSP, MCP, local memory.
AI Security Review
scanned 15d ago · by lpm-firewall-aiCode mode defaults to enabled and evaluates supplied JavaScript. If the optional isolated-vm implementation fails, the fallback uses Node's vm despite an explicit warning that it lacks full process isolation, creating a conditional remote code execution risk without evidence of malicious intent.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
Code mode is enabled and the agent invokes execute_code while isolated-vm is unavailable or fails to load.
Impact
An escape could let supplied code act with the CLI process's authority and bypass the intended tool permission boundary.
Mechanism
The fallback evaluates supplied code with runInNewContext and exposes host callback functions in its context.
AI rationale
Code mode defaults to enabled and evaluates supplied JavaScript. If the optional isolated-vm implementation fails, the fallback uses Node's vm despite an explicit warning that it lacks full process isolation, creating a conditional remote code execution risk without evidence of malicious intent.
Evidence
dist/index.jspackage.json
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The published CLI points to bin/autopilot.mjs, and package imports load dist/index.js.
- Code mode defaults to enabled, and the agent exposes execute_code in that mode.
- The fallback runs supplied code in Node's vm context and exposes host callbacks to it.
- If isolated-vm fails to load, execution falls back to Node's vm; the package warns this fallback lacks full process isolation.
- isolated-vm is an optional dependency, so the fallback can be reached when it is unavailable or fails to load.
Evidence against
- Sandbox tool calls are routed through the configured permission callback. This does not protect against code that escapes the fallback context.
Behavioral surface
ChildProcessCryptoDynamicRequireEnvironmentVarsEvalFilesystemNetworkShell
HighEntropyStringsUrlStrings
Source & flagged code
16 flaggeddist/index.jsView file
41Source sends the broad process environment to a literal external destination.
L41: function defaultLogDir() {
L42: const xdg = process.env.XDG_CONFIG_HOME || join(homedir(), ".config");
L43: return join(xdg, "kimiflare", "logs");
...
L58: if (currentStream) {
L59: currentStream.end();
L60: currentStream = null;
...
L188: function setFetchForTesting(f) {
L189: fetchImpl = f ?? ((url, init) => fetch(url, init));
L190: }
...
L989: import { createSign } from "crypto";
L990: import { execFile } from "child_process";
L991: import { readFile } from "fs/promises";
Critical
Hardcoded Runtime Data Exfiltration
Source sends credentials or rich application records to a package-controlled external receiver enabled by default.
dist/index.jsView on unpkg · L411055const iat = Math.floor(now2 / 1e3);
L1056: const payload = enc({ iss: sa.client_email, scope: SCOPE, aud: sa.token_uri ?? "https://oauth2.googleapis.com/token", iat, exp: iat + 3600 });
L1057: const signature = createSign("RSA-SHA256").update(`${header}.${payload}`).sign(sa.private_key).toString("base64url");
...
L1062: const tokenUrl = creds.type === "service_account" ? creds.token_uri ?? "https://oauth2.googleapis.com/token" : "https://oauth2.googleapis.com/token";
L1063: const body = creds.type === "service_account" ? new URLSearchParams({ grant_type: "urn:ietf:params:oauth:grant-type:jwt-bearer", assertion: signJwt(creds, now2) }) : new URLSearchP...
L1064: const res = await fetchImpl2(tokenUrl, { method: "POST", headers: { "Content-Type": "application/x-www-form-urlencoded" },
Critical
Command Output Exfiltration
Findings
6 Critical9 High4 Medium6 Low
CriticalHardcoded Runtime Data Exfiltrationdist/index.js
CriticalCommand Output Exfiltrationdist/index.js
CriticalDns Exfiltrationdist/index.js
CriticalRemote Response Code Executiondist/index.js
CriticalTrigger Reachable Dangerous Capabilitydist/index.js
CriticalPrevious Version Dangerous Deltadist/sdk/index.d.ts
HighChild Processdist/index.js
HighShelldist/index.js
HighEval
HighSame File Env Network Executiondist/index.js
HighCredential Redirect Persistencedist/index.js
HighRemote Agent Bridgedist/index.js
Affected versions and remediation
This report applies to autopilot-ai@2.2.0.
See all versions of autopilot-ai for other recorded verdicts.
- Review the evidence and your use of autopilot-ai@2.2.0 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.