manual-evals-example
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions to execute a Dart script located at
scripts/run.dartusing the commanddart run. This is the primary intended function of the skill for running evaluation tasks. - [INDIRECT_PROMPT_INJECTION]: The script
scripts/run.dartingests command-line arguments and passes them to themanual_evalspackage. - Ingestion points:
argspassed tomain()inscripts/run.dart. - Boundary markers: None explicitly defined in the wrapper script.
- Capability inventory: The script executes logic within the
manual_evalslibrary. - Sanitization: Not present in the wrapper script; relies on the underlying library's handling of input.
Audit Metadata