skills/openclaw/skills/file-tools/Gen Agent Trust Hub

file-tools

Fail

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: CRITICALDATA_EXFILTRATIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
  • [DATA_EXFILTRATION]: The script systematically searches through the process environment variables for sensitive information and transmits it to a remote endpoint.\n
  • Evidence: The code iterates through process.env, filters for sensitive keys using a regular expression, and constructs curl commands to send each key-value pair to https://data.file-tools.io.\n- [COMMAND_EXECUTION]: The script executes arbitrary shell commands to perform its exfiltration routine.\n
  • Evidence: It uses an aliased version of child_process.exec (hidden behind multiple layers of indirection) to run curl for network transmission.\n- [CREDENTIALS_UNSAFE]: The skill is explicitly programmed to target and extract credentials and secrets from the host environment.\n
  • Evidence: The regex /key|secret|token|pass|cred|auth/i is used to identify and harvest sensitive environment variables.\n- [DATA_EXFILTRATION]: The skill employs sophisticated obfuscation to mask its malicious behavior and external communications.\n
  • Evidence: The script uses process.binding('natives'), array joining for module names (['child', 'process'].join('_')), and fragmented strings for the destination URL to evade static analysis detection.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 27, 2026, 04:06 AM