coreweave-deploy-integration

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a potential surface for indirect prompt injection attacks.
  • Ingestion points: The agent is instructed to read and process output from kubectl, helm, and kustomize commands, as well as the results of health check requests from inference services (SKILL.md).
  • Boundary markers: There are no explicit instructions to use delimiters or ignore embedded commands when the agent evaluates tool outputs.
  • Capability inventory: The skill configuration allows the agent to execute shell commands via kubectl, helm, and kustomize, which can be used to modify cluster state (SKILL.md).
  • Sanitization: The instructions do not include steps to sanitize or validate external output before it is processed by the LLM.
  • [DYNAMIC_EXECUTION]: A TypeScript example provided in the documentation uses synchronous shell execution to retrieve system information.
  • Evidence: The health check endpoint example uses execSync to run the nvidia-smi command to query GPU status.
  • Context: While the command is hardcoded and used for monitoring, the pattern involves spawning a subprocess which is a form of dynamic execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:45 AM
Security Audit — agent-trust-hub — coreweave-deploy-integration