chatting-with-aws-devops-agent

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs the agent to ingest and analyze untrusted local data, including package manifests and IaC files.\n
  • Ingestion points: Local context files like package.json, CDK stacks, and git logs are passed into the message parameter of the aws_devops_agent__chat tool (SKILL.md).\n
  • Boundary markers: Absent. The instructions do not use delimiters or specific framing to separate the untrusted local data from the agent's core instructions.\n
  • Capability inventory: The agent can initiate investigations via the aws_devops_agent__investigate tool and suggests commands for the user to run via the aws CLI.\n
  • Sanitization: No sanitization or validation of the local workspace data is implemented or instructed before the data is processed.\n- [COMMAND_EXECUTION]: The skill provides a fallback mechanism that requires the execution of shell commands using the aws CLI tool to communicate directly with the DevOps Agent service.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 04:00 PM
Security Audit — agent-trust-hub — chatting-with-aws-devops-agent