aws-lambda-durable-functions

Pass

Audited by Gen Agent Trust Hub on Oct 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The troubleshooting agent instructions in references/troubleshooting-executions.md ingest user-supplied Execution ARNs and Regions to perform AWS CLI queries. This creates a potential surface for malformed input processing, though it is mitigated by explicit instructions for read-only diagnosis and the use of built-in CLI query parameters.
  • Ingestion points: references/troubleshooting-executions.md (User-supplied ARN and Region inputs).
  • Boundary markers: Not explicitly defined for variable interpolation.
  • Capability inventory: AWS CLI read operations (aws lambda, aws logs).
  • Sanitization: Not explicitly defined in instructions; relies on agent tool-calling protocol and specified CLI query flags.
  • [EXTERNAL_DOWNLOADS]: The skill guides users to install verified SDKs from the official vendor (awslabs) and trusted organizations (aws) via npm and pip, including @aws/durable-execution-sdk-js and aws-durable-execution-sdk-python.
  • [COMMAND_EXECUTION]: The skill defines a troubleshooting workflow that utilizes AWS CLI commands (get-durable-execution-history, filter-log-events). The instructions include 'CRITICAL SAFETY RULES' that restrict the agent to read-only actions and strictly prohibit termination APIs or state modifications.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 5, 2026, 02:04 PM