The Agent Skills Directory

Standard Data Processing Patterns: The skill is designed to analyze structured data (JSON and CSV files) from a local output directory. It provides clear rules and Python helper functions for common tasks like aggregating metrics, calculating ROI, and grading evidence heterogeneity using well-established libraries like pandas and numpy.
No Sensitive Data Access: The analysis focuses entirely on project-specific data structures related to causal modeling. There are no attempts to access sensitive system files, environment variables, or hardcoded credentials.
Controlled Instructional Environment: The skill uses a structured 'Analysis Playbook' and mandatory statement tagging (Fact/Inference/Hypothesis) to ensure consistent and verifiable agent responses. It includes a trust checklist to help the agent identify and caveat unreliable estimates based on sample size or diagnostic flags.
Absence of Network Operations: The skill operates locally on the provided output directory and does not contain any instructions or code to perform network requests, external downloads, or data exfiltration.
Secure Technical Implementation: The provided Python snippets use safe methods for file handling and data manipulation. No dynamic execution patterns like eval(), exec(), or unsafe deserialization were detected.

causal-insights