autonomous-testing
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill automatically generates test code and fixes based on source code analysis and writes them to the local filesystem. These generated scripts are then executed using testing frameworks (pytest, vitest, playwright). This autonomous 'generate-and-execute' lifecycle represents a dynamic execution risk where the agent runs code it has created at runtime.
- [INDIRECT_PROMPT_INJECTION]: The skill reads project source code to inform AI-driven test generation and code repair. Malicious instructions embedded in the codebase being tested could potentially influence the AI's output.
- Ingestion points: Project source files (*.py, *.ts, *.tsx) and API route definitions are read in Phase 1 and Phase 2.
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are documented for the prompt construction.
- Capability inventory: Execution of shell commands (pytest, npx vitest, npx playwright), filesystem writes (writing generated tests in Phase 2 and applying fixes in Phase 5), and invocation of external binaries (~/bin/deepseek, ~/bin/gemini).
- Sanitization: No sanitization, escaping, or validation of the ingested source code is performed before it is processed by the AI models.
- [COMMAND_EXECUTION]: The skill relies on executing binaries located in the user's home directory (~/bin/deepseek and ~/bin/gemini) rather than standard system-wide or project-local dependencies. This introduces a structural dependency on a non-standard environment state.
Audit Metadata