tilebox-workflow-releases

Pass

Audited by Gen Agent Trust Hub on Nov 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill facilitates the execution of local shell commands via the tilebox CLI to initialize, build, publish, and deploy workflow projects.
  • [EXTERNAL_DOWNLOADS]: The skill manages Python dependencies using uv. It references the official PyTorch CPU wheel repository (download.pytorch.org), which is a well-known technology service used for legitimate package distribution.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a behavior where the agent is authorized to initialize and deploy projects based on "implicit" geospatial prompts.
  • Ingestion points: The agent processes user-provided Earth observation or geospatial prompts (documented in SKILL.md).
  • Boundary markers: The instructions include explicit behavioral boundaries, such as prohibiting production deployments without explicit intent and respecting "local-only" constraints.
  • Capability inventory: The skill provides access to tilebox workflow init, publish-release, and deploy-release (documented across SKILL.md).
  • Sanitization: The skill relies on natural language instructions for the agent to filter intent, rather than programmatically sanitizing the prompts before they influence the deployment logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 2, 2026, 09:06 AM