agents-harden
Pass
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes various bash commands such as 'agentcore status', 'aws iam list-policies', and 'grep' to audit the local project environment and source code. These commands are diagnostic and intended to help the user identify security risks like hardcoded secrets or overly broad IAM roles.
- [PROMPT_INJECTION]: The skill ingests untrusted local data by reading 'agentcore.json' and scanning application source code for patterns. While this constitutes an indirect prompt injection surface (Category 8), the skill is intended for developer use on local projects, and its behavior is confined to security auditing with no network exfiltration.
- [SAFE]: The skill's primary purpose is security hardening. It identifies dangerous platform capabilities like 'InvokeAgentRuntimeCommand' and provides specific guidance on how to secure them. All external URLs point to official AWS documentation or well-known administrative consoles.
Audit Metadata