ft-scout@9.4.0
FrontTerrain — The Onboarding Co-Pilot for New Codebases
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThis is a concrete authentication and unsafe-agent-execution vulnerability, but the dangerous behavior is activated by an explicit CLI workflow rather than an install hook. The scanned email behavior is consistent with signup messaging, not confirmed credential theft.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
Run the Scout CLI, authenticate through Holosene or provide a long direct token, then invoke the agent.
Impact
Commands can read, alter, or delete accessible local data and use the desktop session without per-command approval.
Mechanism
Authentication bypass enabling unrestricted LLM command and computer execution.
AI rationale
This is a concrete authentication and unsafe-agent-execution vulnerability, but the dangerous behavior is activated by an explicit CLI workflow rather than an install hook. The scanned email behavior is consistent with signup messaging, not confirmed credential theft.
Evidence
bin/src/utils/holoseneAuth.jsbin/src/commands/login.jsbin/src/engine/runtime/agentRuntime.jsbin/src/engine/computer/computerController.jsbin/src/engine/computer/scout_driver.exe.ft/auth.json.ft/screenshots
Network endpoints1
scout.frontterrain.com
Decision evidence
public snapshotAI called this Suspicious at 93.0% confidence as Critical Vulnerability with low false-positive risk.
Evidence for warning
- The Holosene login fallback treats any token longer than eight characters as a Pro user.
- Browser login saves the returned tier, enabling the unauthenticated fallback to unlock Pro-only computer control.
- The agent runtime runs an LLM-supplied shell command with no approval or allowlist.
- On Windows, Pro computer control runs a bundled executable and can fall back to encoded PowerShell.
- The package contains no install lifecycle hook, so this behavior requires running its CLI.
Evidence against
- The email code sends account welcome or verification messages to the supplied signup address; no credential exfiltration path was confirmed.
- MCP server setup is an explicit user command and stores its configuration under the package's own .ft directory.
- The shipped tgz is a readable archive and the apparent invoice file contains only mock text.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
Source & flagged code
18 flaggedbin/src/utils/firebaseAuth.jsView file
1patternName = google_api_key
severity = high
line = 1
High
High Secret
Package contains a high-severity secret pattern.
bin/src/utils/firebaseAuth.jsView on unpkg · L11patternName = google_api_key
severity = high
line = 1
High
Secret Pattern
Google API key in bin/src/utils/firebaseAuth.js
bin/src/utils/firebaseAuth.jsView on unpkg · L1bin/src/server/server.jsView file
3import path from 'path';
L4: import { exec } from 'child_process';
L5: import { fileURLToPath } from 'url';
High
Child Process
Package source references child process execution.
bin/src/server/server.jsView on unpkg · L3Findings
3 Critical12 High5 Medium5 Low
CriticalCredential Exfiltrationbin/src/utils/email.js
CriticalTrigger Reachable Dangerous Capabilitybin/src/utils/email.js
CriticalPrevious Version Dangerous Deltabin/src/engine/agentEngine.js
HighHigh Secretbin/src/utils/firebaseAuth.js
HighChild Processbin/src/server/server.js
HighShellbin/src/commands/agent.js
HighSame File Env Network Executionbin/src/utils/holoseneAuth.js
HighSandbox Evasion Gated Capabilitybin/src/commands/setup.js
HighRemote Agent Bridgebin/src/engine/agentEngine.js
HighTrigger Reachable Credential Exfiltrationbin/src/utils/email.js
HighRuntime Package Install
Affected versions and remediation
This report applies to ft-scout@9.4.0.
See all versions of ft-scout for other recorded verdicts.
- Review the evidence and your use of ft-scout@9.4.0 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.