orx-experiment-tree
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill's core functionality revolves around a 'run command', which is a single shell command used to train or evaluate models. The agent is instructed to resolve, record, and execute these commands, which may include compute-specific requirements or backend-specific execution.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources, creating a potential for the agent to be influenced by instructions embedded within that data.
- Ingestion points: The agent reads data from run logs (
orx logs), metrics curves (orx metrics), experiment descriptions (orx exp desc), and Git branches containing project code. - Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions when parsing logs or reviewing code changes.
- Capability inventory: The agent has the capability to execute shell commands (
orx exp run), commit code to Git repositories, and modify experiment metadata. - Sanitization: The skill does not describe any sanitization or validation steps for content retrieved from logs or external repositories before it is used for decision-making in the auto-research loop.
Audit Metadata