The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions facilitate downloading query results and obtaining authentication tokens from verified vendor endpoints at datadata.com. These operations are essential for the platform's functionality and target trusted infrastructure.
[DATA_EXFILTRATION]: Network activity is limited to standard API interactions and the retrieval of user-requested data. The skill mitigates the risk of large-scale data ingestion or unauthorized exfiltration by instructing the agent to save results to temporary local files (/tmp/) and inspect them using localized previews (head/tail), rather than reading complete untrusted datasets into the conversation context.
[PROMPT_INJECTION]: The skill identifies the surface risk of processing external data (e.g., query results) and implements preventative rules, such as prohibiting the automatic selection of data sources and enforcing minimal operation principles where the agent stops and waits for user confirmation after each discrete step.

datadata-manual