platform-engineering

Pass

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill configures an Internal Developer Platform that ingests data from external sources, which creates a potential surface for indirect prompt injection.
  • Ingestion points: The Backstage Software Catalog fetches component definitions from remote GitHub repositories and Scaffolder actions pull template skeletons from external URLs.
  • Boundary markers: The provided configuration examples do not include specific delimiters or instructions for the agent to ignore instructions embedded within ingested YAML or template files.
  • Capability inventory: The platform environment possesses broad capabilities, including the ability to execute shell commands via Scaffolder actions and manage cloud resources through Kubernetes-based control planes (Crossplane).
  • Sanitization: No explicit sanitization or strict validation logic for the content of ingested files is demonstrated in the snippets.
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to fetch and install tools and providers from well-known sources.
  • Backstage: Uses npx to fetch the Backstage application generator from the npm registry.
  • Crossplane: Downloads Helm charts from the official crossplane-stable repository and installs AWS provider packages from Upbound.
  • Score: Installs CLI tools using Homebrew from the official score-spec tap.
  • [SAFE]: The skill demonstrates secure credential management by utilizing environment variables and Kubernetes secrets for sensitive data like GITHUB_TOKEN, database passwords, and service account tokens.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 27, 2026, 04:15 AM