eks-operation-review
Pass
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's configuration fetches and executes MCP servers from the awslabs organization using the uvx tool. These components provide the necessary bridge to AWS and Kubernetes APIs and are recognized as trusted vendor resources.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the EKS environment (such as resource labels, pod metadata, and event messages) into its analysis context.
- Ingestion points: Untrusted data enters the agent context through EKS discovery and Kubernetes resource listing tools specified in .claude/commands/eks-operation-review.md.
- Boundary markers: The skill uses a structured workflow and specific report templates, though it lacks explicit boundary markers or instructions to the agent to ignore embedded commands within ingested resource metadata.
- Capability inventory: The skill has the ability to write files to the local workspace and execute a bundled Python script for report conversion.
- Sanitization: The report conversion tool (tools/report_to_html.py) implements HTML escaping and strict URL scheme validation to mitigate risks during report rendering.
- [COMMAND_EXECUTION]: The skill executes a bundled Python script (tools/report_to_html.py) to convert assessment findings from markdown to HTML. The script is self-contained, uses standard libraries, and includes defensive coding practices to safely handle the report content.
Audit Metadata