The Agent Skills Directory

[PROMPT_INJECTION]: No instructions attempting to override agent behavior or bypass safety guidelines were found. The skill maintains a strict human-approval loop for all outputs.- [CREDENTIALS_UNSAFE]: No hardcoded credentials or secrets. The documentation correctly mandates storing credentials in .env files and accessing them via secure CLI tools.- [EXTERNAL_DOWNLOADS]: All external dependencies are well-known, established data science libraries (e.g., pandas, sf-hamilton, dvc, hydra-core) installed from official package registries.- [REMOTE_CODE_EXECUTION]: No patterns of insecure remote code execution (e.g., curl|bash) or dynamic execution of untrusted code were identified.- [DATA_EXFILTRATION]: No suspicious network activity or indicators of unauthorized data transmission to external domains.- [COMMAND_EXECUTION]: Shell command usage is restricted to legitimate project maintenance tasks, such as running local analysis scripts and invoking official tool CLIs.- [SAFE]: The skill demonstrates high maturity by implementing automated data quality reviews, file hashing for provenance tracking, and clear separation between data acquisition and transformation logic.

analytic-workbench