The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a webhook listener to receive payment status updates from an external API, which creates a potential surface for indirect prompt injection.
Ingestion points: The NotifyURL endpoint (POST /api/webhooks/payuni) defined in SKILL.md and references/code-examples.md processes data from external requests.
Boundary markers: The skill explicitly provides implementation logic for signature verification using HMAC-SHA256 (verifyCheckCode) to authenticate the sender.
Capability inventory: The skill requires Bash, Edit, and Write tools to configure the project and create service files.
Sanitization: The instructions mandate the use of timingSafeEqual (Node.js) or hash_equals (PHP) to validate the CheckCode parameter before the agent or application processes the payload.
[CREDENTIALS_UNSAFE]: No hardcoded secrets were found. The instructions follow security best practices by requiring the agent to search for and guide the user to set sensitive keys (PAYUNI_HASH_KEY, PAYUNI_HASH_IV) within .env configuration files.
[COMMAND_EXECUTION]: The skill suggests using ngrok for local development testing of webhooks, which is a standard industry practice. The command is provided as a documentation example for the user and is not executed autonomously by the agent for malicious purposes.

payuni-checkout