The Agent Skills Directory

[No Code or Execution]: The skill consists entirely of markdown instructions and a metadata configuration file. There are no scripts (Python, Node.js), shell commands, or dynamic context injections present.
[Safe Instruction Design]: The workflow and standards provided focus on logical verification steps, such as checking for survivorship bias, verifying join logic, and ensuring visualization integrity. These are standard best practices for data science and analysis QA.
[Tool Scoping]: The skill references other tools like $jupyter-notebooks and $visualize-data, but these are intended to be invoked by the agent as part of its standard environment rather than through any custom, risky mechanisms.
[Data Handling]: While the skill mentions processing sensitive data like SQL queries or spreadsheets, it does so by providing logical frameworks for human or agent review rather than exfiltrating that data to external services.

validate-data