test-suite-architect
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill utilizes 'Master Prompts' (e.g., in references/master_qa_prompt.md) that employ persona adoption ('senior QA engineer with 20+ years of experience') and authoritative markers ('CRITICAL INSTRUCTIONS', 'MANDATORY RULES') to direct agent behavior and prioritize skill-specific logic over default guidelines.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by instructing the agent to process external data as executable steps.
- Ingestion points: Markdown documentation and test case files (e.g., 02-CLI-TEST-CASES.md).
- Boundary markers: Absent; there are no instructions to delimit untrusted content or ignore embedded malicious instructions within the test cases.
- Capability inventory: The agent is granted the capability to execute shell commands (git, pnpm, docker), write to the file system (CSV updates), and perform environment setup tasks.
- Sanitization: Absent; no validation or escaping logic is provided to ensure that instructions read from test case files are safe to execute.
- [COMMAND_EXECUTION]: The Day 1 onboarding guide and master prompts encourage the execution of various shell commands (e.g., pnpm install, git clone, docker exec) and autonomous execution of steps documented in project files, which could lead to unauthorized system modifications if the source documents are untrusted.
Audit Metadata