@agentvault/openai-bridge@0.1.11
AgentVault OpenAI Codex bridge for encrypted owner messages and rooms
AI Security Review
scanned 6d ago · by lpm-firewall-aiThe source automatically uploads runtime telemetry containing account and conversation metadata, without documenting it or exposing an opt-out. The inspected telemetry does not include message text, and the command-execution alerts are consistent with the bridge’s user-triggered, isolated Codex workflow, so this warrants a warning rather than a block.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
After the authenticated channel is established; buffered spans are flushed automatically every 30 seconds.
Impact
The service receives identifiers and activity metadata, including model usage and conversation details; no message text was found in the telemetry fields inspected.
Mechanism
The telemetry reporter POSTs hub and agent identifiers and buffered usage and reply spans to the AgentVault telemetry endpoint.
AI rationale
The source automatically uploads runtime telemetry containing account and conversation metadata, without documenting it or exposing an opt-out. The inspected telemetry does not include message text, and the command-execution alerts are consistent with the bridge’s user-triggered, isolated Codex workflow, so this warrants a warning rather than a block.
Evidence
dist/index.jsREADME.md
Network endpoints1
api.agentvault.chat/api/v1/telemetry/ingest
Decision evidence
public snapshotAI called this Suspicious at 91.0% confidence as Unknown with medium false-positive risk.
Evidence for warning
- After the authenticated channel is established, the source starts a telemetry reporter that automatically flushes every 30 seconds.
- The reporter posts spans to the configured AgentVault API’s `/api/v1/telemetry/ingest` endpoint and includes hub and agent identifiers.
- The spans contain model and usage metrics plus reply outcome, duration, and conversation metadata. The README does not disclose this runtime telemetry or describe an opt-out.
Evidence against
- The telemetry goes to the AgentVault API, with `api.agentvault.chat` as the default host; no unrelated collector was identified.
- The telemetry fields inspected contain usage and conversation metadata, not message text or Codex credentials.
- Codex execution is user-triggered and configured with read-only sandboxing; the bridge prompt says filesystem, shell, web, and integration tools are unavailable.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
Source & flagged code
9 flaggeddist/index.jsView file
104092await init_bridge();
L104093: import { spawn as spawn2 } from "node:child_process";
L104094: import { realpathSync } from "node:fs";
...
L104104: checkBinary: checkVersion,
L104105: output: (text) => process.stdout.write(text + "\n"),
L104106: command: (binary, args2, options = {}) => new Promise((resolve4, reject) => {
L104107: const child = spawn2(binary, args2, { env: options.env ?? { PATH: "/usr/bin:/bin", HOME: process.env.HOME, XDG_RUNTIME_DIR: process.env.XDG_RUNTIME_DIR, DBUS_SESSION_BUS_ADDRESS: p...
Critical
Command Output Exfiltration
Source executes local commands and sends command output to an external endpoint.
dist/index.jsView on unpkg · L104092104092Trigger-reachable chain: manifest.main -> dist/index.js
L104092: await init_bridge();
L104093: import { spawn as spawn2 } from "node:child_process";
L104094: import { realpathSync } from "node:fs";
...
L104104: checkBinary: checkVersion,
L104105: output: (text) => process.stdout.write(text + "\n"),
L104106: command: (binary, args2, options = {}) => new Promise((resolve4, reject) => {
L104107: const child = spawn2(binary, args2, { env: options.env ?? { PATH: "/usr/bin:/bin", HOME: process.env.HOME, XDG_RUNTIME_DIR: process.env.XDG_RUNTIME_DIR, DBUS_SESSION_BUS_ADDRESS: p...
Critical
Trigger Reachable Dangerous Capability
A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.
dist/index.jsView on unpkg · L104092Findings
2 Critical6 High5 Medium5 Low
CriticalCommand Output Exfiltrationdist/index.js
CriticalTrigger Reachable Dangerous Capabilitydist/index.js
HighChild Processdist/index.js
HighShell
HighSame File Env Network Executiondist/index.js
HighCredential Redirect Persistencedist/index.js
HighTrigger Reachable Command Output Exfiltrationdist/index.js
HighSemantic Analysis Limiteddist/index.js
MediumNetwork
MediumEnvironment Vars
MediumOversized Source Filedist/index.js
MediumOversized Cli Entrypointdist/index.js
MediumStructural Risk Force Deep Review
Affected versions and remediation
This report applies to @agentvault/openai-bridge@0.1.11.
See all versions of @agentvault/openai-bridge for other recorded verdicts.
- Review the evidence and your use of @agentvault/openai-bridge@0.1.11 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.