test-generation

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill is a comprehensive tool for generating unit, integration, and UI tests. It follows platform-specific idioms and promotes deterministic testing practices.\n- [DATA_EXFILTRATION]: The skill includes explicit security constraints, instructing the agent to never include real credentials, tokens, or production data in fixtures, as test data is typically committed to version control.\n- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute standard, project-local test runners and coverage tools (e.g., ./gradlew, flutter test, cargo test, xccov). These commands are standard for the development workflow and are executed within the user's project context.\n- [PROMPT_INJECTION]: The skill processes project source code and existing tests to generate new test files, identifying an indirect prompt injection surface.\n
  • Ingestion points: Project source files and existing test suites read via Read, Grep, and Glob tools.\n
  • Boundary markers: Absent.\n
  • Capability inventory: Includes Bash (command execution), Write (file system modification), and Skill (invoking other agent skills).\n
  • Sanitization: Not mentioned in the instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 05:36 PM
Security Audit — agent-trust-hub — test-generation