prompt-regression-tester
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill framework includes a surface for processing untrusted data through prompt interpolation.
- Ingestion points: Test case inputs and context fields from the JSON definitions in SKILL.md are used to populate prompts.
- Boundary markers: Prompt templates are populated using standard Python string formatting in the comparison framework without specific delimiters to isolate external data.
- Capability inventory: The framework demonstrates the use of an LLM tool to generate outputs for comparison and analysis.
- Sanitization: Input data is used directly as provided in the test case definitions without pre-processing or escaping.
Audit Metadata