The Agent Skills Directory

[SAFE]: The skill implements multiple security best practices for network operations and credential handling.
[DATA_EXPOSURE]: The script explicitly redacts API keys from generated URLs and error messages. It also prevents users from passing credentials as command-line arguments, requiring the use of environment variables (NCBI_API_KEY) which is a standard safe practice for secret management.
[EXTERNAL_DOWNLOADS]: Network requests are strictly limited to a hardcoded whitelist of biological database domains (e.g., rest.uniprot.org, nlm.nih.gov, ebi.ac.uk). The script validates that full URLs provided as arguments match the base URL of the selected service, preventing SSRF and data exfiltration to arbitrary external domains.
[DYNAMIC_EXECUTION]: The skill uses requests and json.loads() for data retrieval and parsing, which are safe methods. It does not utilize dangerous functions like eval() or exec(), and it avoids unsafe deserialization patterns.
[COMMAND_EXECUTION]: The script includes logic to compact and truncate large JSON responses before they are returned to the agent's context, which helps prevent context window flooding or potential prompt injection attacks embedded in large data payloads.

public-db-lookup