jp-validate

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted project files (prd.md, architecture.md) to guide the generation and execution of test code.
  • Ingestion points: Reads documentation from {project-root}/docs/juspay/ (File: steps/step-01-init.md).
  • Capability inventory: Capable of writing files to the codebase and executing shell commands via detected runners or curl/bash (File: steps/step-03-backend-tests.md).
  • Boundary markers: None identified.
  • Sanitization: No explicit sanitization or boundary markers are mentioned for the ingested markdown files.
  • [COMMAND_EXECUTION]: The skill dynamically executes shell commands and test runners based on the project environment.
  • Evidence: Detection and invocation of npm test, pytest, Playwright, Cypress, and curl (File: references/test-stack-detection.md, steps/step-03-backend-tests.md).
  • Mitigation: The skill includes explicit instructions to avoid logging secrets, credentials, or Primary Account Numbers (PANs) during execution (File: references/payment-test-matrix.md).
  • [DATA_EXFILTRATION]: The skill accesses sensitive environmental configuration files.
  • Evidence: Reads .env and secret stores to pass credentials to test processes via environment variables (File: SKILL.md, references/juspay-mcp.md).
  • Context: This access is intended for integration testing and the skill mandates that these values are never echoed or persisted in logs/reports.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 09:39 AM
Security Audit — agent-trust-hub — jp-validate