eval-creator-ci

Pass

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is explicitly designed to execute shell commands defined within evaluation cases. The command-check verification method runs user-defined or generated commands and evaluates their exit codes and output.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface in its 'Create Mode' where it processes external data to generate new executable tests.
  • Ingestion points: The skill reads data from the learning_aggregator_ci workflow artifact or gap reports, which are external to the skill's own code.
  • Boundary markers: No specific delimiters or 'ignore embedded instructions' warnings are mentioned for the processing of these external artifacts.
  • Capability inventory: The skill can execute shell commands (command-check), read and write to the .evals/ directory, and trigger the creation of pull requests via the create-pull-request safe-output.
  • Sanitization: The instructions do not specify any sanitization or validation of the input patterns before they are used to determine verification methods or command strings.
  • [DYNAMIC_EXECUTION]: The skill implements a dynamic code generation and execution pattern. It writes new markdown files to the .evals/cases/ directory containing executable command definitions and subsequently reads and executes those commands in its 'Run Mode'.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 27, 2026, 09:17 PM