aws-agentic-ai

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides the agent with extensive access to AWS CLI toolsets, including sensitive services such as aws secretsmanager and aws s3. These tools allow for the retrieval of credentials and manipulation of data stored in AWS accounts.
  • [EXTERNAL_DOWNLOADS]: The documentation references the installation of standard external packages required for development and validation, including the bedrock-agentcore Python library and @apidevtools/swagger-cli Node.js package.
  • [INDIRECT_PROMPT_INJECTION]: The Browser service (documented in services/browser/README.md) introduces a significant attack surface for tool output poisoning by enabling the extraction of raw content from external websites.
  • Ingestion points: Data enters the agent's context through web content extraction actions like getText and evaluate within the Browser service.
  • Boundary markers: The provided documentation and examples do not implement delimiters or specific instructions to prevent the agent from obeying instructions embedded in scraped web content.
  • Capability inventory: The agent possesses high-impact capabilities, including the ability to manage AWS infrastructure and access secrets via the provided Bash tools.
  • Sanitization: Although the documentation notes 'Sanitize extracted data' as a best practice, the skill does not provide or enforce automated sanitization logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 07:57 PM