aws-agentic-ai
Warn
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill redirects the agent to fetch and process numerous external documentation files from the GitHub repository
github.com/zxkane/aws-skills. This creates a dependency on content maintained by an unverified community member rather than official AWS sources. - [REMOTE_CODE_EXECUTION]: The documentation includes links to several remote Python and Docker templates (e.g.,
runtime-fastapi-template.py,Dockerfile.runtime-template) and instructs the agent to read and implement them. These files are hosted in a public community repository and are not subject to official security vetting. - [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection due to its reliance on external data ingestion. 1. Ingestion points: The agent is instructed to read multiple remote README and template files from the
zxkane/aws-skillsrepository. 2. Boundary markers: The instructions do not define clear delimiters or safeguards to prevent the agent from following malicious instructions that might be embedded within those external files. 3. Capability inventory: The skill is designed to generate and execute AWS CLI commands, manage S3 storage, and deploy Lambda functions and containerized runtimes. 4. Sanitization: There is no evidence of validation or sanitization procedures to verify the safety of the remote code templates before implementation.
Audit Metadata