mitm-find-idor
Pass
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use shell commands such as
mitmdump,grep, andcurlfor traffic analysis and vulnerability verification. - [DATA_EXFILTRATION]: The provided methodology includes using
curlwith session cookies to test external endpoints, which involves transmitting sensitive authentication data over the network. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted network traffic data from
log.txt. 1. Ingestion points: The agent reads captured traffic fromlog.txt. 2. Boundary markers: None are specified to separate log data from agent instructions. 3. Capability inventory: The skill utilizes shell commands (grep,curl) that could be targeted by injected content. 4. Sanitization: No validation of log content is performed before processing.
Audit Metadata