troubleshooting-application-failures

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [Ingestion of Untrusted Data]: The skill is designed to search and analyze CloudWatch log groups and streams. Since log content is generated by external applications and can be influenced by external actors, it represents an indirect prompt injection surface. The agent processes this data to identify error patterns and stack traces. While this is inherent to the skill's purpose, it is a point where malicious instructions could theoretically be embedded in log messages to influence the agent's analysis.
  • [AWS Command Execution]: The troubleshooting procedure involves executing various AWS CLI commands (e.g., describe-log-groups, describe-log-streams, start-query) using the call_aws tool. These commands are parameterized by user-provided inputs such as application_name and region. The skill relies on the agent's execution environment to safely handle these parameters and prevent command injection.
  • [Privilege Configuration]: The documentation correctly identifies and lists the specific IAM permissions required for the skill to operate (e.g., logs:DescribeLogGroups, logs:StartQuery). This follows the principle of least privilege by informing the user of the necessary access scope rather than requesting administrative rights.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 11:58 AM
Security Audit — agent-trust-hub — troubleshooting-application-failures