qa-team
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the
gadugi-agentic-testframework from the author's GitHub repository and uses the official GitHub API (api.github.com) to perform version freshness checks. These activities are transparent, well-documented, and targeted at established service endpoints. - [COMMAND_EXECUTION]: To achieve its goal of behavioral testing, the skill performs various command-line operations, such as launching binaries, executing shell scripts within "shadow environments," and interacting with terminal interfaces. These operations are essential to the skill's primary function as a QA tool.
- [DYNAMIC_EXECUTION]: The skill interprets and executes test logic defined in declarative YAML scenarios at runtime. This allows the agent to perform sequences of actions like application launches, inputs, and validations based on external configuration files.
- [INDIRECT_PROMPT_INJECTION]: The skill utilizes AI agents to interpret raw application outputs (such as stdout, HTML content, or screenshots) to verify test success. This represents an indirect prompt injection surface where malicious output from a tested application could theoretically attempt to influence the reasoning agent.
- Ingestion points: Application output (text/images) and YAML scenario files.
- Boundary markers: Standard YAML structure; however, there are no explicit instructions for agents to disregard instructions embedded within the application output itself.
- Capability inventory: Subprocess launching, browser automation, file system modification, and network interactions.
- Sanitization: No explicit sanitization or filtering of application output before processing by comprehension agents is mentioned.
Audit Metadata