vastai-local-dev-loop
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill follows standard development best practices, such as recommending the use of .env files for local secrets management and providing mock objects for unit testing to avoid live API calls during development.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a workflow where the agent manages local project files and executes shell commands. This represents a potential surface for indirect prompt injection if the agent were to process untrusted data within those files or command outputs, though no specific malicious implementation is present.
- Ingestion points: Output from the
vastaiCLI tool, local source code and test files invastai-project/. - Boundary markers: Not explicitly defined in the provided templates.
- Capability inventory: Bash execution (docker, vastai, pip), File writing and editing.
- Sanitization: No explicit sanitization of command output is performed within the provided script examples.
Audit Metadata