The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various development and testing commands (e.g., mvn test, npm install, npx playwright test) through its qa_agent.py script. This is the primary function of the tool and is used to automate the testing process of the target repository.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md instructions utilize the !command`` syntax to dynamically determine the installation path of the skill using a find command. This is used for self-discovery and environment setup, which is a benign and standard practice for this platform.
[EXTERNAL_DOWNLOADS]: The skill can trigger the installation of necessary testing runtimes and browsers (e.g., @playwright/test) from official registries like NPM. These operations target well-known and trusted services, following the platform's security guidelines.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository (source code, diffs) to generate test cases and perform risk analysis. While this presents an attack surface where malicious code comments could influence the agent, the skill implements a mandatory human confirmation step for all generated test cases, effectively mitigating this risk. Severity for this category is low as it identifies a vulnerability surface rather than active exploitation.
[SAFE]: The tool accesses local configuration files and .env files to retrieve necessary test environment variables. It includes explicit instructions to avoid printing or exfiltrating sensitive information like passwords or tokens, ensuring data access is limited to the intended local testing scope.

quality-assurance-agent