test-software-behavior
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data, including source code, test configurations, and fixtures, which creates an attack surface for instructions embedded in processed data to influence agent behavior.
- Ingestion points: Reads existing tests, test configuration, implementation details, and visible fixtures in SKILL.md and references/regressions-and-nondeterminism.md.
- Boundary markers: None identified to delimit external code from instructions.
- Capability inventory: The skill facilitates command execution via test runners and file modifications for adding tests as described in SKILL.md and references/generative-and-mutation.md.
- Sanitization: No explicit sanitization of input code or data is mentioned.
- [EXTERNAL_DOWNLOADS]: The instructions reference the use and installation of several external, well-known testing frameworks and tools.
- Evidence: Mentions Hypothesis (hypothesis.readthedocs.io), Stryker (stryker-mutator.io), Playwright (playwright.dev), and LLVM libFuzzer (llvm.org).
- Context: These references are informative and point to official documentation for standard industry tools, following best practices for software testing.
Audit Metadata