The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill performs network requests to the official Ctrip API endpoint (https://externalcallback.ctrip.com). This is legitimate behavior for a travel service and targets a well-known service provider domain.
[CREDENTIALS_UNSAFE]: The skill manages an API token (WENDAO_API_KEY) for authentication. It appropriately recommends the use of environment variables to store the secret and provides explicit instructions for users to avoid hardcoding or publicly sharing their keys.
[INDIRECT_PROMPT_INJECTION]: The skill processes user input and API output, which represents a potential attack surface. However, the instructions direct the agent to specifically extract only the result field from the structured API response, reducing the likelihood of the agent following instructions that might be embedded in raw response data.
[PROMPT_INJECTION]: The core requirements include instructions to prioritize the skill's API responses over the agent's general knowledge. While this involves behavior override patterns, they are standard for domain-specific assistant skills to ensure data accuracy.

ctrip-wendao