input-configurations

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill provides templates for runtime script execution used by the Elastic Agent to process data.
  • references/aws-s3-guide.md includes a JavaScript template for the sqs.notification_parsing_script.source field, which parses SQS notifications.
  • references/http-endpoint-guide.md, references/websocket-guide.md, and references/httpjson-guide.md describe the use of CEL (Common Expression Language) and Go templates for runtime data transformation and request logic.- [INDIRECT_PROMPT_INJECTION]: The skill defines configurations for various inputs that ingest data from external, untrusted sources, creating a potential attack surface for indirect injection if the processed data is later used in LLM prompts.
  • Ingestion points: Multiple files (e.g., references/aws-s3-guide.md, references/http-endpoint-guide.md, references/tcp-udp-guide.md) describe ingesting data from cloud notifications, buckets, webhooks, and network listeners.
  • Boundary markers: The skill recommends using Handlebars variables and structured configuration but does not explicitly define delimiters for separating untrusted data from agent instructions.
  • Capability inventory: The agent inputs have capabilities for network operations, file reading, and executing internal parsing logic (JS/CEL).
  • Sanitization: The guides recommend using redact.fields for sensitive data and provide patterns for structured JSON parsing.- [SAFE]: The skill is authored by the vendor (Elastic) and exclusively references official documentation and vendor-controlled infrastructure (e.g., elastic.co, elastic-cspm-cft.s3.eu-central-1.amazonaws.com). It proactively warns users against hardcoding credentials and provides patterns for secure authentication.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 10:59 AM