debugging

Fail

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides code to initialize rpdb.set_trace(), which opens a remote debugging session on a network port. This allows any user with network access to the port to execute arbitrary Python code in the execution environment.
  • [DATA_EXFILTRATION]: The debug_state implementation creates a Flask route that returns comprehensive system and process information, including memory usage, CPU statistics, process IDs, and environment versions. Exposing this metadata via a web endpoint is a significant information disclosure risk.
  • [CREDENTIALS_UNSAFE]: The NetworkDebugger class is designed to log the entirety of request and response headers. This practice routinely captures sensitive authentication material such as 'Authorization' tokens, 'Cookie' headers, and 'X-API-Key' values, leading to credential exposure in logs.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 18, 2026, 08:57 AM
Security Audit — agent-trust-hub — debugging