testing-load-balancers
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION] (MEDIUM): The skill explicitly reads from '{baseDir}/config/', a directory likely to contain sensitive infrastructure details, environment secrets, or credentials. This access increases the risk of sensitive data exposure.
- [COMMAND_EXECUTION] (MEDIUM): The skill uses a Bash tool restricted by the pattern 'test:loadbalancer-*'. While limited, this pattern allows execution of local scripts or binaries that could be abused if an attacker gains the ability to place or modify files matching the wildcard.
- [PROMPT_INJECTION] (MEDIUM): Risk of Indirect Prompt Injection exists because the skill ingests untrusted data (configuration files and test outputs) and uses it to 'generate insights' and 'detect patterns'. If these sources contain malicious instructions, the agent may follow them using its Write or Bash tools.
Audit Metadata