valohai-yaml-step
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to generate shell commands for environment setup and script execution within a valohai.yaml configuration file, including apt-get and pip installations.- [EXTERNAL_DOWNLOADS]: The skill encourages defining inputs from external sources like S3 buckets and downloading packages from public registries.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It directs the agent to analyze local project files (e.g., requirements.txt, train.py) to determine configuration parameters and command logic. Malicious instructions embedded in these analyzed files could influence the generated YAML or helper scripts. Ingestion points: Project scripts and dependency files analyzed in the 'Analyze the Project' section. Boundary markers: None specified for the analysis of project files. Capability inventory: File creation (valohai.yaml, helper Python scripts) and shell command generation. Sanitization: No instructions provided for sanitizing or validating content extracted from project files.
Audit Metadata