The Agent Skills Directory

[SAFE]: The skill defines a structured analytical workflow using reputable open-source libraries such as pandas, polars, marimo, and Hamilton.
[CREDENTIALS_UNSAFE]: No hardcoded credentials or unsafe secret handling was detected. The instructions explicitly direct users to store sensitive information in .env files, which is a standard security best practice.
[DATA_EXFILTRATION]: No malicious data exfiltration patterns were found. Network operations mentioned (e.g., DVC remotes, MLflow tracking) are standard for the described data science use cases and are configured by the user.
[REMOTE_CODE_EXECUTION]: The skill does not contain any patterns for executing untrusted remote code. Documentation for package installation and environment setup refers exclusively to well-known and trusted software packages from official registries.
[COMMAND_EXECUTION]: CLI commands provided in templates (such as mkdir, pip install, and dvc init) are routine operations for project scaffolding and development environment management.
[PROMPT_INJECTION]: The skill uses structured declarations (Plan, Run, Review, Summary) to organize agent behavior and reporting, but these do not contain instructions to bypass safety filters or override core agent constraints.

analytic-workbench