aws-rightsizing
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill defines a large number of AWS CLI, jq, and kubectl commands for the agent to execute in order to collect resource inventory and CloudWatch performance metrics. The instructions include security practices such as verifying the target account ID and using temporary kubeconfig files for Kubernetes interactions.
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes data from AWS resource tags and metadata.
- Ingestion points: Untrusted data enters the context through resource names and tags collected via AWS CLI describe commands into inventory JSON files.
- Boundary markers: There are no explicit delimiters or markers used in the Python logic (core.py) to separate data from instructions.
- Capability inventory: The agent is granted capabilities to run shell commands and read AWS environment data based on the skill logic.
- Sanitization: No specific sanitization or filtering of external metadata is performed before the data is processed by core.py or included in final reports.
- [DYNAMIC_EXECUTION]: The skill uses jq templates to dynamically construct complex MetricDataQuery JSON structures at runtime, which are then used in AWS CloudWatch API calls via the agent's shell environment.
Audit Metadata