byted-supabase

Pass

Audited by Gen Agent Trust Hub on Nov 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources such as database query outputs, storage listings, and AI Gateway responses. This data is ingested into the agent's context and used to generate explanations or decide on subsequent actions, creating an attack surface for indirect prompt injection.
  • Ingestion points: Outputs from db query, storage ls, functions list, ai-gateway, and sandbox stdout/stderr are ingested into the conversation flow.
  • Boundary markers: The instructions emphasize verification of work and adherence to security checklists, providing defensive context, though they do not mandate the use of strict delimiters for all ingested data.
  • Capability inventory: The agent can execute SQL queries, manage file systems, deploy serverless functions, and execute code within a sandboxed environment.
  • Sanitization: The skill provides extensive documentation for users on how to implement sanitization and RLS in their own applications to mitigate data-driven risks.
  • [EXTERNAL_DOWNLOADS]: The skill examples and documentation reference external dependencies and code from the vendor's official GitHub repository.
  • Evidence: The Python example in pyproject.toml references byted-supabase-sandbox hosted at github.com/volcengine/byted-supabase-cli.git.
  • [COMMAND_EXECUTION]: The skill interacts with the host environment and remote instances using the byted-supabase-cli tool and sandboxed shell commands.
  • Evidence: The skill body in SKILL.md and various references describe using byted-supabase-cli for infrastructure management tasks.
  • [DYNAMIC_EXECUTION]: The skill provides a feature to create and execute code within isolated Linux containers (Sandboxes) to perform tasks like data processing or agent execution.
  • Evidence: The sandbox commands allow for dynamic execution of Python, Node.js, and shell scripts within isolated environments using the base and code-interpreter templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 12, 2026, 06:34 AM