The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands, specifically curl and bash, as a fallback when no established test runner is detected. It also invokes existing project runners like npm test, pytest, and playwright test.
[REMOTE_CODE_EXECUTION]: The skill dynamically generates and persists test scripts (e.g., Playwright specs, Python test files) within the user's codebase, which are subsequently executed to validate the integration.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its reliance on project-level documentation to drive test generation logic.
Ingestion points: Reads architecture.md, prd.md, task-checklist.md, and handoff files in step-01-init.md.
Boundary markers: No explicit delimiters or 'ignore embedded instructions' warnings are defined for these data sources.
Capability inventory: The skill has permissions to write files to the project directory and execute shell commands and test runners.
Sanitization: There is no evidence of filtering or sanitizing the logic extracted from documentation files before it is used to generate executable code.
[DATA_EXFILTRATION]: The skill transmits telemetry data via the juspay_track_integration_stage tool. This data includes the integration phase, status, product type, and platform. Although the instructions prohibit the inclusion of sensitive data or PII in the metadata, the tool serves as a persistent network egress point to the vendor's infrastructure.

jp-validate